Android MediaCodecVideoDecoder: Manage lifetime of texture frames

talk/app/webrtc/java/jni/androidmediadecoder_jni.cc

Index: talk/app/webrtc/java/jni/androidmediadecoder_jni.cc
diff --git a/talk/app/webrtc/java/jni/androidmediadecoder_jni.cc b/talk/app/webrtc/java/jni/androidmediadecoder_jni.cc
index dce5a22e8262ef3ee2329b09d85d63553714e294..0b2c6a7a3072e1717c3de8536a9cb0e10bfd646e 100644
--- a/talk/app/webrtc/java/jni/androidmediadecoder_jni.cc
+++ b/talk/app/webrtc/java/jni/androidmediadecoder_jni.cc
@@ -33,6 +33,7 @@
 #include "talk/app/webrtc/java/jni/androidmediacodeccommon.h"
 #include "talk/app/webrtc/java/jni/classreferenceholder.h"
 #include "talk/app/webrtc/java/jni/native_handle_impl.h"
+#include "talk/app/webrtc/java/jni/surfacetexturehelper_jni.h"
 #include "webrtc/base/bind.h"
 #include "webrtc/base/checks.h"
 #include "webrtc/base/logging.h"
@@ -111,7 +112,7 @@ class MediaCodecVideoDecoder : public webrtc::VideoDecoder,
   bool use_surface_;
   VideoCodec codec_;
   webrtc::I420BufferPool decoded_frame_pool_;
-  NativeHandleImpl native_handle_;
+  rtc::scoped_refptr<SurfaceTextureHelper> surface_texture_helper_;
   DecodedImageCallback* callback_;
   int frames_received_;  // Number of frames received by decoder.
   int frames_decoded_;  // Number of frames decoded by decoder.
@@ -144,10 +145,11 @@ class MediaCodecVideoDecoder : public webrtc::VideoDecoder,
   jfieldID j_height_field_;
   jfieldID j_stride_field_;
   jfieldID j_slice_height_field_;
-  jfieldID j_surface_texture_field_;
   // MediaCodecVideoDecoder.DecodedTextureBuffer fields.
   jfieldID j_textureID_field_;
-  jfieldID j_texture_presentation_timestamp_us_field_;
+  jfieldID j_transform_matrix_field_;
+  jfieldID j_texture_timestamp_ns_field_;
+  jfieldID j_texture_decode_time_ms_field_;
   // MediaCodecVideoDecoder.DecodedByteBuffer fields.
   jfieldID j_info_index_field_;
   jfieldID j_info_offset_field_;
@@ -156,8 +158,6 @@ class MediaCodecVideoDecoder : public webrtc::VideoDecoder,
 
   // Global references; must be deleted in Release().
   std::vector<jobject> input_buffers_;
-  jobject surface_texture_;
-  jobject previous_surface_texture_;
 
   // Render EGL context - owned by factory, should not be allocated/destroyed
   // by VideoDecoder.
@@ -171,8 +171,6 @@ MediaCodecVideoDecoder::MediaCodecVideoDecoder(
     key_frame_required_(true),
     inited_(false),
     sw_fallback_required_(false),
-    surface_texture_(NULL),
-    previous_surface_texture_(NULL),
     codec_thread_(new Thread()),
     j_media_codec_video_decoder_class_(
         jni,
@@ -191,7 +189,7 @@ MediaCodecVideoDecoder::MediaCodecVideoDecoder(
   j_init_decode_method_ = GetMethodID(
       jni, *j_media_codec_video_decoder_class_, "initDecode",
       "(Lorg/webrtc/MediaCodecVideoDecoder$VideoCodecType;"
-      "IILandroid/opengl/EGLContext;)Z");
+      "IILorg/webrtc/SurfaceTextureHelper;)Z");
   j_release_method_ =
       GetMethodID(jni, *j_media_codec_video_decoder_class_, "release", "()V");
   j_dequeue_input_buffer_method_ = GetMethodID(
@@ -221,17 +219,17 @@ MediaCodecVideoDecoder::MediaCodecVideoDecoder(
       jni, *j_media_codec_video_decoder_class_, "stride", "I");
   j_slice_height_field_ = GetFieldID(
       jni, *j_media_codec_video_decoder_class_, "sliceHeight", "I");
-  j_surface_texture_field_ = GetFieldID(
-      jni, *j_media_codec_video_decoder_class_, "surfaceTexture",
-      "Landroid/graphics/SurfaceTexture;");
 
   jclass j_decoder_decoded_texture_buffer_class = FindClass(jni,
       "org/webrtc/MediaCodecVideoDecoder$DecodedTextureBuffer");
   j_textureID_field_ = GetFieldID(
       jni, j_decoder_decoded_texture_buffer_class, "textureID", "I");
-  j_texture_presentation_timestamp_us_field_ =
-      GetFieldID(jni, j_decoder_decoded_texture_buffer_class,
-                 "presentationTimestampUs", "J");
+  j_transform_matrix_field_ = GetFieldID(
+      jni, j_decoder_decoded_texture_buffer_class, "transformMatrix", "[F");
+  j_texture_timestamp_ns_field_ = GetFieldID(
+      jni, j_decoder_decoded_texture_buffer_class, "timestampNs", "J");
+  j_texture_decode_time_ms_field_ = GetFieldID(
+      jni, j_decoder_decoded_texture_buffer_class, "decodeTimeMs", "J");
 
   jclass j_decoder_decoded_byte_buffer_class = FindClass(jni,
       "org/webrtc/MediaCodecVideoDecoder$DecodedByteBuffer");
@@ -254,14 +252,6 @@ MediaCodecVideoDecoder::MediaCodecVideoDecoder(
 MediaCodecVideoDecoder::~MediaCodecVideoDecoder() {
   // Call Release() to ensure no more callbacks to us after we are deleted.
   Release();
-  // Delete global references.
-  JNIEnv* jni = AttachCurrentThreadIfNeeded();
-  if (previous_surface_texture_ != NULL) {
-    jni->DeleteGlobalRef(previous_surface_texture_);
-  }
-  if (surface_texture_ != NULL) {
-    jni->DeleteGlobalRef(surface_texture_);
-  }
 }
 
 int32_t MediaCodecVideoDecoder::InitDecode(const VideoCodec* inst,
@@ -311,6 +301,11 @@ int32_t MediaCodecVideoDecoder::InitDecodeOnCodecThread() {
   frames_received_ = 0;
   frames_decoded_ = 0;
 
+  if (use_surface_) {
+    surface_texture_helper_ = new rtc::RefCountedObject<SurfaceTextureHelper>(
+        jni, render_egl_context_);
+  }
+
   jobject j_video_codec_enum = JavaEnumFromIndex(
       jni, "MediaCodecVideoDecoder$VideoCodecType", codecType_);
   bool success = jni->CallBooleanMethod(
@@ -319,7 +314,8 @@ int32_t MediaCodecVideoDecoder::InitDecodeOnCodecThread() {
       j_video_codec_enum,
       codec_.width,
       codec_.height,
-      use_surface_ ? render_egl_context_ : nullptr);
+      use_surface_ ? surface_texture_helper_->GetJavaSurfaceTextureHelper()
+                   : nullptr);
   if (CheckException(jni) || !success) {
     ALOGE << "Codec initialization error - fallback to SW codec.";
     sw_fallback_required_ = true;
@@ -361,15 +357,6 @@ int32_t MediaCodecVideoDecoder::InitDecodeOnCodecThread() {
     }
   }
 
-  if (use_surface_) {
-    jobject surface_texture = GetObjectField(
-        jni, *j_media_codec_video_decoder_, j_surface_texture_field_);
-    if (previous_surface_texture_ != NULL) {
-      jni->DeleteGlobalRef(previous_surface_texture_);
-    }
-    previous_surface_texture_ = surface_texture_;
-    surface_texture_ = jni->NewGlobalRef(surface_texture);
-  }
   codec_thread_->PostDelayed(kMediaCodecPollMs, this);
 
   return WEBRTC_VIDEO_CODEC_OK;
@@ -394,6 +381,7 @@ int32_t MediaCodecVideoDecoder::ReleaseOnCodecThread() {
   }
   input_buffers_.clear();
   jni->CallVoidMethod(*j_media_codec_video_decoder_, j_release_method_);
+  surface_texture_helper_ = nullptr;
   inited_ = false;
   rtc::MessageQueueManager::Clear(this);
   if (CheckException(jni)) {
@@ -501,9 +489,9 @@ int32_t MediaCodecVideoDecoder::DecodeOnCodecThread(
   // Try to drain the decoder and wait until output is not too
   // much behind the input.
   if (frames_received_ > frames_decoded_ + max_pending_frames_) {
-    ALOGV("Received: %d. Decoded: %d. Wait for output...",
-        frames_received_, frames_decoded_);
-    if (!DeliverPendingOutputs(jni, kMediaCodecTimeoutMs * 1000)) {
+    ALOGD << "Received: " << frames_received_ << ". Decoded: "
+          << frames_decoded_ << ". Wait for output...";
+    if (!DeliverPendingOutputs(jni, kMediaCodecTimeoutMs)) {
       ALOGE << "DeliverPendingOutputs error";
       return ProcessHWErrorOnCodecThread();
     }
@@ -566,7 +554,7 @@ int32_t MediaCodecVideoDecoder::DecodeOnCodecThread(
 }
 
 bool MediaCodecVideoDecoder::DeliverPendingOutputs(
-    JNIEnv* jni, int dequeue_timeout_us) {
+    JNIEnv* jni, int dequeue_timeout_ms) {
   if (frames_received_ <= frames_decoded_) {
     // No need to query for output buffers - decoder is drained.
     return true;
@@ -575,7 +563,7 @@ bool MediaCodecVideoDecoder::DeliverPendingOutputs(
   jobject j_decoder_output_buffer = jni->CallObjectMethod(
       *j_media_codec_video_decoder_,
       j_dequeue_output_buffer_method_,
-      dequeue_timeout_us);
+      dequeue_timeout_ms);
   if (CheckException(jni)) {
     ALOGE << "dequeueOutputBuffer() error";
     return false;
@@ -596,18 +584,25 @@ bool MediaCodecVideoDecoder::DeliverPendingOutputs(
 
   rtc::scoped_refptr<webrtc::VideoFrameBuffer> frame_buffer;
   long output_timestamps_ms = 0;
+  int decode_time_ms = 0;
   if (use_surface_) {
     // Extract data from Java DecodedTextureBuffer.
     const int texture_id =
         GetIntField(jni, j_decoder_output_buffer, j_textureID_field_);
-    const int64_t timestamp_us =
-        GetLongField(jni, j_decoder_output_buffer,
-                     j_texture_presentation_timestamp_us_field_);
-    output_timestamps_ms = timestamp_us / rtc::kNumMicrosecsPerMillisec;
+    const jfloatArray j_transform_matrix =
+        reinterpret_cast<jfloatArray>(GetObjectField(
+            jni, j_decoder_output_buffer, j_transform_matrix_field_));
+    const int64_t timestamp_ns = GetLongField(jni, j_decoder_output_buffer,
+                                              j_texture_timestamp_ns_field_);
+    output_timestamps_ms = timestamp_ns / rtc::kNumNanosecsPerMillisec;
+
+    decode_time_ms = GetLongField(jni, j_decoder_output_buffer,
+                                  j_texture_decode_time_ms_field_);
+    output_timestamps_ms = timestamp_ns / rtc::kNumNanosecsPerMillisec;

[magjed_webrtc, 2015/10/28 11:57:16]

This is a duplicated line.

[perkj_webrtc, 2015/10/28 21:12:39]

Done.

+
     // Create webrtc::VideoFrameBuffer with native texture handle.
-    native_handle_.SetTextureObject(surface_texture_, texture_id);
-    frame_buffer = new rtc::RefCountedObject<JniNativeHandleBuffer>(
-        &native_handle_, width, height);
+    frame_buffer = surface_texture_helper_->CreateTextureFrame(
+        width, height, NativeHandleImpl(jni, texture_id, j_transform_matrix));
   } else {
     // Extract data from Java ByteBuffer and create output yuv420 frame -
     // for non surface decoding only.
@@ -692,23 +687,24 @@ bool MediaCodecVideoDecoder::DeliverPendingOutputs(
     decoded_frame.set_ntp_time_ms(ntp_times_ms_.front());
     ntp_times_ms_.erase(ntp_times_ms_.begin());
   }
-  int64_t frame_decoding_time_ms = 0;
+  int64_t frame_delayed_ms = 0;
   if (frame_rtc_times_ms_.size() > 0) {
-    frame_decoding_time_ms = GetCurrentTimeMs() - frame_rtc_times_ms_.front();
+    frame_delayed_ms = GetCurrentTimeMs() - frame_rtc_times_ms_.front();
     frame_rtc_times_ms_.erase(frame_rtc_times_ms_.begin());
   }
   ALOGV("Decoder frame out # %d. %d x %d. %d x %d. Color: 0x%x. TS: %ld."
-      " DecTime: %lld", frames_decoded_, width, height, stride, slice_height,
-      color_format, output_timestamps_ms, frame_decoding_time_ms);
+      " DelayTime: %lld", frames_decoded_, width, height, stride, slice_height,

[magjed_webrtc, 2015/10/28 11:57:16]

I think you should print both latency and decode time.

[perkj_webrtc, 2015/10/28 21:12:39]

Done.

+      color_format, output_timestamps_ms, frame_delayed_ms);
 
   // Calculate and print decoding statistics - every 3 seconds.
   frames_decoded_++;
   current_frames_++;
-  current_decoding_time_ms_ += frame_decoding_time_ms;
+  current_decoding_time_ms_ += decode_time_ms;

[magjed_webrtc, 2015/10/28 11:57:16]

You need to set |decode_time_ms| for ByteBuffer output as well. You can probably
use |frame_delayed_ms|.

[perkj_webrtc, 2015/10/28 21:12:39]

Done.

   int statistic_time_ms = GetCurrentTimeMs() - start_time_ms_;
   if (statistic_time_ms >= kMediaCodecStatisticsIntervalMs &&
       current_frames_ > 0) {
-    ALOGD << "Decoded frames: " << frames_decoded_ << ". Bitrate: " <<
+    ALOGD << "Decoded frames: " << frames_decoded_ <<  ". Received frames: "
+        <<  frames_received_ << ".  Bitrate: " <<
         (current_bytes_ * 8 / statistic_time_ms) << " kbps, fps: " <<
         ((current_frames_ * 1000 + statistic_time_ms / 2) / statistic_time_ms)
         << ". decTime: " << (current_decoding_time_ms_ / current_frames_) <<
@@ -720,7 +716,8 @@ bool MediaCodecVideoDecoder::DeliverPendingOutputs(
   }
 
   // Callback - output decoded frame.
-  const int32_t callback_status = callback_->Decoded(decoded_frame);
+  const int32_t callback_status =
+      callback_->Decoded(decoded_frame, decode_time_ms);
   if (callback_status > 0) {
     ALOGE << "callback error";
   }