Add support for direct submission of buffers to the renderers without a separate thread

moonlight-common/src/com/limelight/nvstream/av/audio/AudioDepacketizer.java

@@ -11,9 +11,16 @@ public class AudioDepacketizer {
 	private LinkedBlockingQueue<ByteBufferDescriptor> decodedUnits =
 			new LinkedBlockingQueue<ByteBufferDescriptor>(DU_LIMIT);
 	
+	private AudioRenderer directSubmitRenderer;
+	
 	// Sequencing state
 	private short lastSequenceNumber;
 	
+	public AudioDepacketizer(AudioRenderer directSubmitRenderer)
+	{
+		this.directSubmitRenderer = directSubmitRenderer;
+	}
+
 	private void decodeData(byte[] data, int off, int len)
 	{
 		// Submit this data to the decoder
@@ -24,8 +31,10 @@ public class AudioDepacketizer {
 			// Return value of decode is frames (shorts) decoded per channel
 			decodeLen *= 2*OpusDecoder.getChannelCount();
 			
-			// Put it on the decoded queue
-			if (!decodedUnits.offer(new ByteBufferDescriptor(pcmData, 0, decodeLen))) {
+			if (directSubmitRenderer != null) {
+				directSubmitRenderer.playDecodedAudio(pcmData, 0, decodeLen);
+			}
+			else if (!decodedUnits.offer(new ByteBufferDescriptor(pcmData, 0, decodeLen))) {
 				System.out.println("Audio player too slow! Forced to drop decoded samples");
 				// Clear out the queue
 				decodedUnits.clear();

moonlight-common/src/com/limelight/nvstream/av/audio/AudioRenderer.java

@@ -1,6 +1,11 @@
 package com.limelight.nvstream.av.audio;
 
 public interface AudioRenderer {
+	// playDecodedAudio() is lightweight, so don't use an extra thread for playback
+	public static final int CAPABILITY_DIRECT_SUBMIT = 0x1;
+	
+	public int getCapabilities();
+	
 	public void streamInitialized(int channelCount, int sampleRate);
 	
 	public void playDecodedAudio(byte[] audioData, int offset, int length);

moonlight-common/src/com/limelight/nvstream/av/audio/AudioStream.java

@@ -23,7 +23,7 @@ public class AudioStream {
 	
 	private DatagramSocket rtp;
 	
-	private AudioDepacketizer depacketizer = new AudioDepacketizer();
+	private AudioDepacketizer depacketizer;
 	
 	private LinkedList<Thread> threads = new LinkedList<Thread>();
 	
@@ -79,7 +79,9 @@ public class AudioStream {
 		
 		startDepacketizerThread();
 		
+		if ((streamListener.getCapabilities() & AudioRenderer.CAPABILITY_DIRECT_SUBMIT) == 0) {
 			startDecoderThread();
+		}
 		
 		startUdpPingThread();
 	}
@@ -102,6 +104,13 @@ public class AudioStream {
 		}
 		
 		streamListener.streamInitialized(OpusDecoder.getChannelCount(), OpusDecoder.getSampleRate());
+		
+		if ((streamListener.getCapabilities() & AudioRenderer.CAPABILITY_DIRECT_SUBMIT) != 0) {
+			depacketizer = new AudioDepacketizer(streamListener);
+		}
+		else {
+			depacketizer = new AudioDepacketizer(null);
+		}
 	}
 	
 	private void startDepacketizerThread()

moonlight-common/src/com/limelight/nvstream/av/video/VideoDecoderRenderer.java

@@ -7,6 +7,11 @@ public interface VideoDecoderRenderer {
 	public static final int FLAG_FORCE_HARDWARE_DECODING = 0x2;
 	public static final int FLAG_FORCE_SOFTWARE_DECODING = 0x4;
 	
+	// SubmitDecodeUnit() is lightweight, so don't use an extra thread for decoding
+	public static final int CAPABILITY_DIRECT_SUBMIT = 0x1;
+	
+	public int getCapabilities();
+
 	public void setup(int width, int height, int redrawRate, Object renderTarget, int drFlags);
 	
 	public void start();

moonlight-common/src/com/limelight/nvstream/av/video/VideoDepacketizer.java

@@ -22,12 +22,14 @@ public class VideoDepacketizer {
 	private ByteBufferDescriptor cachedDesc = new ByteBufferDescriptor(null, 0, 0);
 	
 	private ConnectionStatusListener controlListener;
+	private VideoDecoderRenderer directSubmitDr;
 	
-	private static final int DU_LIMIT = 7;
+	private static final int DU_LIMIT = 15;
 	private LinkedBlockingQueue<DecodeUnit> decodedUnits = new LinkedBlockingQueue<DecodeUnit>(DU_LIMIT);
 	
-	public VideoDepacketizer(ConnectionStatusListener controlListener)
+	public VideoDepacketizer(VideoDecoderRenderer directSubmitDr, ConnectionStatusListener controlListener)
 	{
+		this.directSubmitDr = directSubmitDr;
 		this.controlListener = controlListener;
 	}
 	
@@ -43,7 +45,11 @@ public class VideoDepacketizer {
 		if (avcNalDataChain != null && avcNalDataLength != 0) {
 			// Construct the H264 decode unit
 			DecodeUnit du = new DecodeUnit(DecodeUnit.TYPE_H264, avcNalDataChain, avcNalDataLength, 0);
-			if (!decodedUnits.offer(du)) {
+			if (directSubmitDr != null) {
+				// Submit directly to the decoder
+				directSubmitDr.submitDecodeUnit(du);
+			}
+			else if (!decodedUnits.offer(du)) {
 				// We need a new IDR frame since we're discarding data now
 				System.out.println("Video decoder is too slow! Forced to drop decode units");
 				decodedUnits.clear();

moonlight-common/src/com/limelight/nvstream/av/video/VideoStream.java

@@ -35,6 +35,7 @@ public class VideoStream {
 	private LinkedList<Thread> threads = new LinkedList<Thread>();
 
 	private NvConnectionListener listener;
+	private ConnectionStatusListener avConnListener;
 	private VideoDepacketizer depacketizer;
 	private StreamConfiguration streamConfig;
 	
@@ -47,7 +48,7 @@ public class VideoStream {
 	{
 		this.host = host;
 		this.listener = listener;
-		this.depacketizer = new VideoDepacketizer(avConnListener);
+		this.avConnListener = avConnListener;
 		this.streamConfig = streamConfig;
 	}
 	
@@ -134,6 +135,13 @@ public class VideoStream {
 		if (decRend != null) {
 			decRend.setup(streamConfig.getWidth(), streamConfig.getHeight(),
 					60, renderTarget, drFlags);
+			
+			if ((decRend.getCapabilities() & VideoDecoderRenderer.CAPABILITY_DIRECT_SUBMIT) != 0) {
+				depacketizer = new VideoDepacketizer(decRend, avConnListener);
+			}
+			else {
+				depacketizer = new VideoDepacketizer(null, avConnListener);
+			}
 		}
 	}
 
@@ -164,8 +172,10 @@ public class VideoStream {
 			// Start the depacketizer thread to deal with the RTP data
 			startDepacketizerThread();
 			
-			// Start decoding the data we're receiving
+			// Start a decode thread if we're not doing direct submit
+			if ((decRend.getCapabilities() & VideoDecoderRenderer.CAPABILITY_DIRECT_SUBMIT) == 0) {
 				startDecoderThread();
+			}
 			
 			// Start the renderer
 			decRend.start();