Fix temporal SVC reference picture model for VP9Enc

Now the 2 layer temporal model is similar to libvpx mode.

eg: libvpx:
./vp9_spatial_svc_encoder --width=1280 --height=720 --passes=1 --rc-end-usage=1 --target-bitrate=4096 --min-bitrate=4096 --max-bitrate=4096 --spatial-layers=1 --temporal-layers=2 --temporal-layering-mode=2 --timebase=1/30 --kf-dist=30 -o sample_vp9.webm /home/sreerenj/Videos/raw/sample_1280_720_300f_i420.yuv

eg: gst:gst-launch-1.0 videotestsrc num-buffers=300 ! video/x-raw, width=1280, height=720, framerate=30/1 ! vaapivp9enc enable-software-brc=true bitrate=4096 k-svc=false temporal-svc=true ! webmmux ! filesink location=sample.webm

1 files changed, 59 insertions, 25 deletions

diff --git a/gst-libs/gst/vaapi/gstvaapiencoder_vp9.c b/gst-libs/gst/vaapi/gstvaapiencoder_vp9.c
index fbd8b856..798c7f1f 100644
--- a/gst-libs/gst/vaapi/gstvaapiencoder_vp9.c
+++ b/gst-libs/gst/vaapi/gstvaapiencoder_vp9.c
@@ -30,6 +30,8 @@
 #include "gstvaapisurface.h"
 #include "gstvaapifilter.h"
 
+#define GST_VP9_REF_FRAMES_TEMPORAL_SVC 2
+
 #if USE_LIBMEBO
 #include <libmebo/libmebo.h>
 #endif
@@ -406,24 +408,37 @@ get_ref_indices_k_svc (guint ref_list_idx, guint spatial_group_count,
 }
 
 static void
-get_ref_indices_temporal_svc (guint ref_list_idx, guint temporal_id,
+get_ref_indices_temporal_svc (guint frame_num, guint ref_list_idx, guint temporal_id,
     guint * last_idx, guint * gf_idx, guint * arf_idx,
     guint * ref_frame_ctrl_l0, guint8 * refresh_frame_flags)
 {
   gint last_filled_idx;
 
-  last_filled_idx = (ref_list_idx - 1) & (GST_VP9_REF_FRAMES - 1);
+  last_filled_idx = (ref_list_idx - 1) % GST_VP9_REF_FRAMES_TEMPORAL_SVC;
   
   *ref_frame_ctrl_l0 = 0x07;
-  *refresh_frame_flags = 1 << ((last_filled_idx + 1) % GST_VP9_REF_FRAMES);
+
+  if (frame_num % 2 != 0)
+    *refresh_frame_flags = 2;
+  else
+    *refresh_frame_flags = 1;
+
+  //g_message ("Debug: frame_num=%d get_ref_indices last-filled-idx= %d temporal_id=%d refresh_frame_flags=%d",frame_num, last_filled_idx, temporal_id,*refresh_frame_flags);
 
   //Non Base Layer
   if (temporal_id == 1) {
-    *last_idx = last_filled_idx;
-    *arf_idx = *gf_idx = (last_filled_idx - 1) & (GST_VP9_REF_FRAMES - 1);
+    *last_idx = *gf_idx = last_filled_idx;
+    if (frame_num == 1) {
+      *arf_idx =  *last_idx;
+    } else { 
+      if (last_filled_idx == 0)
+        *arf_idx = 1;
+      else
+        *arf_idx = 0;
+    }
   } else if (temporal_id == 0) {
     //Base Layer
-    *last_idx = *arf_idx = *gf_idx = ((last_filled_idx-1)  & (GST_VP9_REF_FRAMES - 1));
+    *last_idx = *arf_idx = *gf_idx = 0;;
   } else {
     g_message ("Temproal id is >1, something went wrong! ");
   }
@@ -480,7 +495,29 @@ fill_picture_temporal_svc (GstVaapiEncoderVP9 * encoder,
   guint8 refresh_frame_flags = 0;
   guint width, height;
   guint ref_frame_ctrl_l0 = 0;
+  guint i =0;
+  /* Update Reference Frame list */
+  if (picture->type == GST_VAAPI_PICTURE_TYPE_I)
+    memset (pic_param->reference_frames, 0xFF,
+        sizeof (pic_param->reference_frames));
+  else {
+    for (i = 0; i < G_N_ELEMENTS (pic_param->reference_frames); i++) {
+      if (encoder->ref_list[i])
+        pic_param->reference_frames[i] =
+            GST_VAAPI_SURFACE_PROXY_SURFACE_ID (encoder->ref_list[i]);
+      else
+        pic_param->reference_frames[i] = VA_INVALID_ID;
+    }
+  }
 
+  pic_param->luma_ac_qindex = encoder->yac_qi;
+  pic_param->luma_dc_qindex_delta = 1;
+  pic_param->chroma_ac_qindex_delta = 1;
+  pic_param->chroma_dc_qindex_delta = 1;
+  pic_param->filter_level = encoder->loop_filter_level;
+  pic_param->sharpness_level = encoder->sharpness_level;
+
+  pic_param->pic_flags.bits.show_frame = 1;
   width =
       gst_vaapi_surface_get_width (GST_VAAPI_SURFACE_PROXY_SURFACE (surface));
   height =
@@ -497,7 +534,7 @@ fill_picture_temporal_svc (GstVaapiEncoderVP9 * encoder,
   if (picture->type == GST_VAAPI_PICTURE_TYPE_P) {
     pic_param->pic_flags.bits.frame_type = GST_VP9_INTER_FRAME;
 
-    get_ref_indices_temporal_svc (encoder->ref_list_idx, picture->temporal_id,
+    get_ref_indices_temporal_svc (encoder->frame_num-1,encoder->ref_list_idx, picture->temporal_id,
         &last_idx, &gf_idx, &arf_idx, &ref_frame_ctrl_l0, &refresh_frame_flags);
 
     pic_param->ref_flags.bits.ref_frame_ctrl_l0 = ref_frame_ctrl_l0;
@@ -524,6 +561,10 @@ fill_picture (GstVaapiEncoderVP9 * encoder,
   pic_param->reconstructed_frame = GST_VAAPI_SURFACE_PROXY_SURFACE_ID (surface);
   pic_param->coded_buf = GST_VAAPI_CODED_BUFFER_ID (codedbuf);
 
+  //temporal-svc
+  if (encoder->temporal_svc && (encoder->num_temporal_layers))
+    return fill_picture_temporal_svc (encoder, picture, codedbuf, surface, pic_param);
+
   /* Update Reference Frame list */
   if (picture->type == GST_VAAPI_PICTURE_TYPE_I)
     memset (pic_param->reference_frames, 0xFF,
@@ -548,9 +589,6 @@ fill_picture (GstVaapiEncoderVP9 * encoder,
   if (encoder->k_svc && (encoder->num_spatial_layers))
     return fill_picture_k_svc (encoder, picture, codedbuf, surface, pic_param);
 
-  //temporal-svc
-  if (encoder->temporal_svc && (encoder->num_temporal_layers))
-    return fill_picture_temporal_svc (encoder, picture, codedbuf, surface, pic_param);
 
   /* It is possible to have dynamic scaling with gpu by providing
    * src and destination resoltuion. For now we are just using
@@ -658,22 +696,11 @@ update_ref_list_temporal_svc (GstVaapiEncoderVP9 * encoder,
     GstVaapiEncPicture * picture, GstVaapiSurfaceProxy * ref)
 {
   guint i;
-
-  if (picture->type == GST_VAAPI_PICTURE_TYPE_I) {
-    for (i = 0; i < G_N_ELEMENTS (encoder->ref_list); i++)
-      gst_vaapi_surface_proxy_replace (&encoder->ref_list[i], ref);
-    gst_vaapi_surface_proxy_unref (ref);
-    /* set next free slot index */
-    encoder->ref_list_idx = 1;
-    return;
-  } else {
-    if (encoder->ref_list_idx == 0)
-      encoder->ref_list_idx = 1;
-  }
   i = encoder->ref_list_idx;
   gst_vaapi_surface_proxy_replace (&encoder->ref_list[i], ref);
   gst_vaapi_surface_proxy_unref (ref);
-  encoder->ref_list_idx = (encoder->ref_list_idx + 1) % GST_VP9_REF_FRAMES;
+  encoder->ref_list_idx = (encoder->ref_list_idx + 1) % GST_VP9_REF_FRAMES_TEMPORAL_SVC;
+  //g_message ("Debug: update_ref_list: ref_list_idx=next_free_slot = %d",encoder->ref_list_idx);
 }
 
 static GstVaapiEncoderStatus
@@ -813,7 +840,10 @@ gst_vaapi_encoder_vp9_encode (GstVaapiEncoder * base_encoder,
 
     encoder->loop_filter_level = lf;
 
-    g_message ("QP = %d : LF_Strength = %d",qp,lf);
+    if (picture->type != GST_VAAPI_PICTURE_TYPE_P)
+      g_message ("KeyFrame: QP = %d : LF_Strength = %d",qp,lf);
+    else
+      g_message ("PFrame: QP = %d : LF_Strength = %d",qp,lf);
   }
 #endif
 
@@ -843,7 +873,10 @@ gst_vaapi_encoder_vp9_encode (GstVaapiEncoder * base_encoder,
   if (encoder->software_brc) {
     gint32 buf_size;
     buf_size = gst_vaapi_coded_buffer_get_size (GST_VAAPI_CODED_BUFFER_PROXY_BUFFER (codedbuf));
-    g_message ("PostEncodeBufferSize = %d ",buf_size);
+    if (picture->type != GST_VAAPI_PICTURE_TYPE_P)
+      g_message ("KeyFrame: PostEncodeBufferSize = %d ",buf_size);
+    else
+      g_message ("PFrame: PostEncodeBufferSize = %d ",buf_size);
 
     libmebo_rate_controller_update_frame_size (encoder->libmebo_rc, buf_size);
   }
@@ -1021,6 +1054,7 @@ gst_vaapi_encoder_vp9_reordering_temporal_svc (GstVaapiEncoder * base_encoder,
 
   picture->temporal_id = encoder->temporal_id;
   encoder->temporal_id = (encoder->temporal_id + 1) % encoder->num_temporal_layers;
+  //g_message ("Debug:Reordering_TemporalId = %d \n",picture->temporal_id);
   return status;
 }
 static GstVaapiEncoderStatus