Support double pcm encoding in extractor

libraries/common/src/main/java/androidx/media3/common/util/Util.java

@@ -2421,6 +2421,25 @@ public static Format getPcmFormat(AudioProcessor.AudioFormat audioFormat) {
     }
   }
 
+  /**
+   * Converts a sample bit depth to a corresponding little-endian float PCM encoding constant.
+   *
+   * @param bitDepth The bit depth. Supported values are 32 and 64.
+   * @return The corresponding float PCM encoding. If the bit depth is unsupported then {@link
+   *     C#ENCODING_INVALID} is returned.
+   */
+  @UnstableApi
+  public static @C.PcmEncoding int getFloatPcmEncoding(int bitDepth) {
+    switch (bitDepth) {
+      case 32:
+        return C.ENCODING_PCM_FLOAT;
+      case 64:
+        return C.ENCODING_PCM_DOUBLE;
+      default:
+        return C.ENCODING_INVALID;
+    }
+  }
+
   /**
    * Returns whether {@code encoding} is one of the linear PCM encodings.
    *

libraries/exoplayer/src/main/java/androidx/media3/exoplayer/audio/ToFloatPcmAudioProcessor.java

@@ -28,7 +28,9 @@
  * encodings are supported as input:
  *
  * <ul>
+ *   <li>{@link C#ENCODING_PCM_8BIT}
  *   <li>{@link C#ENCODING_PCM_16BIT}
+ *   <li>{@link C#ENCODING_PCM_16BIT_BIG_ENDIAN}
  *   <li>{@link C#ENCODING_PCM_24BIT}
  *   <li>{@link C#ENCODING_PCM_24BIT_BIG_ENDIAN}
  *   <li>{@link C#ENCODING_PCM_32BIT}
@@ -47,7 +49,7 @@ public final class ToFloatPcmAudioProcessor extends BaseAudioProcessor {
   public AudioFormat onConfigure(AudioFormat inputAudioFormat)
       throws UnhandledAudioFormatException {
     @C.PcmEncoding int encoding = inputAudioFormat.encoding;
-    if (!Util.isEncodingHighResolutionPcm(encoding) && encoding != C.ENCODING_PCM_16BIT) {
+    if (!Util.isEncodingLinearPcm(encoding)) {
       throw new UnhandledAudioFormatException(inputAudioFormat);
     }
     return encoding != C.ENCODING_PCM_FLOAT
@@ -64,6 +66,13 @@ public void queueInput(ByteBuffer inputBuffer) {
 
     ByteBuffer buffer;
     switch (inputAudioFormat.encoding) {
+      case C.ENCODING_PCM_8BIT:
+        buffer = replaceOutputBuffer(size * 4);
+        for (int i = position; i < limit; i += 2) {
+          int pcm32BitInteger = ((inputBuffer.get(i) & 0xFF) << 24);
+          writePcm32BitFloat(pcm32BitInteger, buffer);
+        }
+        break;
       case C.ENCODING_PCM_16BIT:
         buffer = replaceOutputBuffer(size * 2);
         for (int i = position; i < limit; i += 2) {
@@ -72,6 +81,14 @@ public void queueInput(ByteBuffer inputBuffer) {
           writePcm32BitFloat(pcm32BitInteger, buffer);
         }
         break;
+      case C.ENCODING_PCM_16BIT_BIG_ENDIAN:
+        buffer = replaceOutputBuffer(size * 2);
+        for (int i = position; i < limit; i += 2) {
+          int pcm32BitInteger =
+              ((inputBuffer.get(i + 1) & 0xFF) << 16) | ((inputBuffer.get(i) & 0xFF) << 24);
+          writePcm32BitFloat(pcm32BitInteger, buffer);
+        }
+        break;
       case C.ENCODING_PCM_24BIT:
         buffer = replaceOutputBuffer((size / 3) * 4);
         for (int i = position; i < limit; i += 3) {
@@ -120,8 +137,6 @@ public void queueInput(ByteBuffer inputBuffer) {
           buffer.putFloat((float) inputBuffer.getDouble(i));
         }
         break;
-      case C.ENCODING_PCM_8BIT:
-      case C.ENCODING_PCM_16BIT_BIG_ENDIAN:
       case C.ENCODING_PCM_FLOAT:
       case C.ENCODING_INVALID:
       case Format.NO_VALUE:

libraries/exoplayer/src/test/java/androidx/media3/exoplayer/audio/ToFloatPcmAudioProcessorTest.java

@@ -42,12 +42,14 @@ public class ToFloatPcmAudioProcessorTest {
    *   <li>{@link C#ENCODING_PCM_16BIT}
    *   <li>{@link C#ENCODING_PCM_32BIT}
    *   <li>{@link C#ENCODING_PCM_24BIT}
+   *   <li>{@link C#ENCODING_PCM_8BIT}
    *   <li>{@link C#ENCODING_PCM_32BIT_BIG_ENDIAN}
    *   <li>{@link C#ENCODING_PCM_24BIT_BIG_ENDIAN}
+   *   <li>{@link C#ENCODING_PCM_16BIT_BIG_ENDIAN}
    *   <li>{@link C#ENCODING_PCM_DOUBLE}
    * </ul>
    */
-  @TestParameter({"2", "22", "21", "1610612736", "1342177280", "1879048192"})
+  @TestParameter({"2", "22", "21", "3", "1610612736", "1342177280", "268435456", "1879048192"})
   private int pcmEncoding;
 
   @Test
@@ -87,7 +89,10 @@ public void configure_returnsFloatPcmEncoding() throws Exception {
 
   private static float getToleranceForEncoding(int pcmEncoding) {
     switch (pcmEncoding) {
+      case C.ENCODING_PCM_8BIT:
+        return 1f / 0x80;
       case C.ENCODING_PCM_16BIT:
+      case C.ENCODING_PCM_16BIT_BIG_ENDIAN:
         return 1f / 0x8000;
       case C.ENCODING_PCM_32BIT:
       case C.ENCODING_PCM_32BIT_BIG_ENDIAN:
@@ -107,9 +112,20 @@ private static float getToleranceForEncoding(int pcmEncoding) {
    */
   private static ByteBuffer getTestSamplesForEncoding(int pcmEncoding) {
     switch (pcmEncoding) {
+      case C.ENCODING_PCM_8BIT:
+        return createByteBuffer(
+            new short[] {Byte.MAX_VALUE, Byte.MIN_VALUE, Byte.MAX_VALUE / 2, Byte.MIN_VALUE / 2});
       case C.ENCODING_PCM_16BIT:
         return createByteBuffer(
             new short[] {Short.MAX_VALUE, Short.MIN_VALUE, 0x4000, (short) 0xC000});
+      case C.ENCODING_PCM_16BIT_BIG_ENDIAN:
+        return createByteBuffer(
+            new short[] {
+              Short.reverseBytes(Short.MAX_VALUE),
+              Short.reverseBytes(Short.MIN_VALUE),
+              Short.reverseBytes((short) 0x4000),
+              Short.reverseBytes((short) 0xC000)
+            });
       case C.ENCODING_PCM_32BIT:
         return createByteBuffer(
             new int[] {

libraries/exoplayer/src/test/java/androidx/media3/exoplayer/e2etest/Mp4PlaybackTest.java

@@ -70,6 +70,7 @@ public static ImmutableList<Sample> mediaSamples() {
         Sample.forFile("sample_opus_fragmented.mp4"),
         Sample.forFile("sample_opus.mp4"),
         Sample.forFile("sample_alac.mp4"),
+        Sample.forFile("sample_fpcm_64le.mp4"),
         Sample.forFile("sample_partially_fragmented.mp4"),
         Sample.withSubtitles("sample_with_vobsub.mp4", "eng"),
         Sample.forFile("testvid_1022ms.mp4"),

libraries/extractor/src/main/java/androidx/media3/extractor/WavUtil.java

@@ -76,6 +76,7 @@ public static int getTypeForPcmEncoding(@C.PcmEncoding int pcmEncoding) {
       case C.ENCODING_PCM_32BIT:
         return TYPE_PCM;
       case C.ENCODING_PCM_FLOAT:
+      case C.ENCODING_PCM_DOUBLE:
         return TYPE_FLOAT;
       // TYPE_PCM is little endian so big endian formats don't match.
       case C.ENCODING_PCM_16BIT_BIG_ENDIAN:
@@ -98,7 +99,7 @@ public static int getTypeForPcmEncoding(@C.PcmEncoding int pcmEncoding) {
       case TYPE_WAVE_FORMAT_EXTENSIBLE:
         return Util.getPcmEncoding(bitsPerSample);
       case TYPE_FLOAT:
-        return bitsPerSample == 32 ? C.ENCODING_PCM_FLOAT : C.ENCODING_INVALID;
+        return Util.getFloatPcmEncoding(bitsPerSample);
       default:
         return C.ENCODING_INVALID;
     }

libraries/extractor/src/main/java/androidx/media3/extractor/mkv/MatroskaExtractor.java

@@ -2451,9 +2451,8 @@ public void initializeFormat(int trackId) throws ParserException {
           break;
         case CODEC_ID_PCM_FLOAT:
           mimeType = MimeTypes.AUDIO_RAW;
-          if (audioBitDepth == 32) {
-            pcmEncoding = C.ENCODING_PCM_FLOAT;
-          } else {
+          pcmEncoding = Util.getFloatPcmEncoding(audioBitDepth);
+          if (pcmEncoding == C.ENCODING_INVALID) {
             pcmEncoding = Format.NO_VALUE;
             mimeType = MimeTypes.AUDIO_UNKNOWN;
             Log.w(

libraries/extractor/src/main/java/androidx/media3/extractor/mp4/BoxParser.java

@@ -2131,8 +2131,8 @@ private static void parseAudioSampleEntry(
       boolean isBigEndian = (formatSpecificFlags & (1 << 1)) != 0;
       if (!isFloat) {
         pcmEncoding = Util.getPcmEncoding(bitsPerSample, isBigEndian ? BIG_ENDIAN : LITTLE_ENDIAN);
-      } else if (!isBigEndian && bitsPerSample == 32) {
-        pcmEncoding = C.ENCODING_PCM_FLOAT;
+      } else if (!isBigEndian) {
+        pcmEncoding = Util.getFloatPcmEncoding(bitsPerSample);
       }
       if (pcmEncoding == C.ENCODING_INVALID) {
         pcmEncoding = Format.NO_VALUE;
@@ -2389,11 +2389,9 @@ private static void parseAudioSampleEntry(
         int sampleSize = parent.readUnsignedByte();
         if (atomType == Mp4Box.TYPE_ipcm) {
           pcmEncoding = Util.getPcmEncoding(sampleSize, byteOrder);
-        } else if (atomType == Mp4Box.TYPE_fpcm
-            && sampleSize == 32
-            && byteOrder.equals(LITTLE_ENDIAN)) {
-          // Only single-width little-endian floating point PCM is supported.
-          pcmEncoding = C.ENCODING_PCM_FLOAT;
+        } else if (atomType == Mp4Box.TYPE_fpcm && byteOrder.equals(LITTLE_ENDIAN)) {
+          // Only little-endian floating point PCM is supported.
+          pcmEncoding = Util.getFloatPcmEncoding(sampleSize);
         }
         if (pcmEncoding != Format.NO_VALUE) {
           mimeType = MimeTypes.AUDIO_RAW;

libraries/extractor/src/test/java/androidx/media3/extractor/mkv/MatroskaExtractorTest.java

@@ -243,6 +243,14 @@ public void mkvSample_withDtsX() throws Exception {
         simulationConfig);
   }
 
+  @Test
+  public void mkaSample_withFpcm64le() throws Exception {
+    ExtractorAsserts.assertBehavior(
+        getExtractorFactory(subtitlesParsedDuringExtraction),
+        "media/mka/bear-pcm-f64le.mka",
+        simulationConfig);
+  }
+
   private static ExtractorAsserts.ExtractorFactory getExtractorFactory(
       boolean subtitlesParsedDuringExtraction) {
     SubtitleParser.Factory subtitleParserFactory;

libraries/extractor/src/test/java/androidx/media3/extractor/mp4/Mp4ExtractorParameterizedTest.java

@@ -296,6 +296,11 @@ public void mp4SampleWith32leFpcm() throws Exception {
     assertExtractorBehavior("media/mp4/sample_fpcm_32le.mp4", /* peekLimit= */ 50);
   }
 
+  @Test
+  public void mp4SampleWith64leFpcm() throws Exception {
+    assertExtractorBehavior("media/mp4/sample_fpcm_64le.mp4", /* peekLimit= */ 50);
+  }
+
   // Only the rotation part of the transformation matrix is resolved (b/390422593 tracks supporting
   // reflection too).
   @Test

libraries/extractor/src/test/java/androidx/media3/extractor/wav/WavExtractorTest.java

@@ -74,4 +74,10 @@ public void sample_wav_format_extensible() throws Exception {
     ExtractorAsserts.assertBehavior(
         WavExtractor::new, "media/wav/sample_wav_format_extensible.wav", simulationConfig);
   }
+
+  @Test
+  public void sample_float64() throws Exception {
+    ExtractorAsserts.assertBehavior(
+        WavExtractor::new, "media/wav/sample_float64.wav", simulationConfig);
+  }
 }

libraries/test_data/src/test/assets/extractordumps/mka/bear-pcm-f64le.mka.0.dump

@@ -0,0 +1,137 @@
+seekMap:
+  isSeekable = true
+  duration = 2740000
+  getPosition(0) = [[timeUs=0, position=517]]
+  getPosition(1) = [[timeUs=0, position=517]]
+  getPosition(1370000) = [[timeUs=0, position=517]]
+  getPosition(2740000) = [[timeUs=0, position=517]]
+numberOfTracks = 1
+track 1:
+  total output bytes = 2104320
+  sample count = 29
+  format 0:
+    id = 1
+    containerMimeType = video/x-matroska
+    sampleMimeType = audio/raw
+    channelCount = 2
+    sampleRate = 48000
+    pcmEncoding = 1879048192
+    selectionFlags = [default]
+    language = und
+  sample 0:
+    time = 0
+    flags = 1
+    data = length 73728, hash 2996BA53
+  sample 1:
+    time = 96000
+    flags = 1
+    data = length 73728, hash AD09C69
+  sample 2:
+    time = 192000
+    flags = 1
+    data = length 73728, hash EFF0EB35
+  sample 3:
+    time = 288000
+    flags = 1
+    data = length 73728, hash 37139E5
+  sample 4:
+    time = 384000
+    flags = 1
+    data = length 73728, hash 45490039
+  sample 5:
+    time = 480000
+    flags = 1
+    data = length 73728, hash B17D13CD
+  sample 6:
+    time = 576000
+    flags = 1
+    data = length 73728, hash 7EC01A5F
+  sample 7:
+    time = 672000
+    flags = 1
+    data = length 73728, hash 8F6D115F
+  sample 8:
+    time = 768000
+    flags = 1
+    data = length 73728, hash 938D8753
+  sample 9:
+    time = 864000
+    flags = 1
+    data = length 73728, hash 15F052F5
+  sample 10:
+    time = 960000
+    flags = 1
+    data = length 73728, hash 859E73FD
+  sample 11:
+    time = 1056000
+    flags = 1
+    data = length 73728, hash D7382BED
+  sample 12:
+    time = 1152000
+    flags = 1
+    data = length 73728, hash FB136FDD
+  sample 13:
+    time = 1248000
+    flags = 1
+    data = length 73728, hash 26DC3CD9
+  sample 14:
+    time = 1344000
+    flags = 1
+    data = length 73728, hash 7935274D
+  sample 15:
+    time = 1440000
+    flags = 1
+    data = length 73728, hash 90EBEB55
+  sample 16:
+    time = 1536000
+    flags = 1
+    data = length 73728, hash 4C9C2FDD
+  sample 17:
+    time = 1632000
+    flags = 1
+    data = length 73728, hash A686C633
+  sample 18:
+    time = 1728000
+    flags = 1
+    data = length 73728, hash D555EFBF
+  sample 19:
+    time = 1824000
+    flags = 1
+    data = length 73728, hash 2082D5C5
+  sample 20:
+    time = 1920000
+    flags = 1
+    data = length 73728, hash D0A9421D
+  sample 21:
+    time = 2016000
+    flags = 1
+    data = length 73728, hash 6625DD21
+  sample 22:
+    time = 2112000
+    flags = 1
+    data = length 73728, hash E05A3C81
+  sample 23:
+    time = 2208000
+    flags = 1
+    data = length 73728, hash 25191701
+  sample 24:
+    time = 2304000
+    flags = 1
+    data = length 73728, hash 65E0F955
+  sample 25:
+    time = 2400000
+    flags = 1
+    data = length 73728, hash 335038B5
+  sample 26:
+    time = 2496000
+    flags = 1
+    data = length 73728, hash 3DBF4769
+  sample 27:
+    time = 2592000
+    flags = 1
+    data = length 73728, hash 3032129C
+  sample 28:
+    time = 2688000
+    flags = 1
+    data = length 39936, hash 63D01025
+tracksEnded = true