libs: decoder: update reference list for SCC.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer-vaapi/-/merge_requests/311>

1 files changed, 39 insertions, 1 deletions

diff --git a/gst-libs/gst/vaapi/gstvaapidecoder_h265.c b/gst-libs/gst/vaapi/gstvaapidecoder_h265.c
index f6be666e..015c870e 100644
--- a/gst-libs/gst/vaapi/gstvaapidecoder_h265.c
+++ b/gst-libs/gst/vaapi/gstvaapidecoder_h265.c
@@ -1345,7 +1345,21 @@ decode_current_picture (GstVaapiDecoderH265 * decoder)
 
   priv->decoder_state |= sps_pi->state;
   if (!(priv->decoder_state & GST_H265_VIDEO_STATE_GOT_I_FRAME)) {
-    if (priv->decoder_state & GST_H265_VIDEO_STATE_GOT_P_SLICE)
+    const GstH265PPS *pps = get_pps (decoder);
+    /* 7.4.3.3.3: the picture is an IRAP picture, nuh_layer_id is equal to 0,
+       and pps_curr_pic_ref_enabled_flag is equal to 0, slice_type shall be
+       equal to 2(I Slice).
+       And F.8.3.4: Decoding process for reference picture lists construction
+       is invoked at the beginning of the decoding process for each P or B
+       slice.
+       so if pps_curr_pic_ref_enabled_flag is set, which means the picture can
+       ref to itself, the IRAP picture may be set to P/B slice, in order to
+       generate the ref lists. If the slice_type is I, no ref list will be
+       constructed and no MV data for that slice according to the syntax.
+       That kind of CVS may start with P/B slice, but in fact it is a intra
+       frame. */
+    if (priv->decoder_state & GST_H265_VIDEO_STATE_GOT_P_SLICE &&
+        !pps->pps_scc_extension_params.pps_curr_pic_ref_enabled_flag)
       goto drop_frame;
     sps_pi->state |= GST_H265_VIDEO_STATE_GOT_I_FRAME;
   }
@@ -1670,6 +1684,7 @@ init_picture_refs (GstVaapiDecoderH265 * decoder,
   guint num_ref_idx_l0_active_minus1 = 0;
   guint num_ref_idx_l1_active_minus1 = 0;
   GstH265RefPicListModification *ref_pic_list_modification;
+  GstH265PPS *const pps = get_pps (decoder);
   guint type;
 
   memset (priv->RefPicList0, 0, sizeof (GstVaapiPictureH265 *) * 16);
@@ -1702,6 +1717,8 @@ init_picture_refs (GstVaapiDecoderH265 * decoder,
     for (i = 0; i < priv->NumPocLtCurr && rIdx < NumRpsCurrTempList0;
         rIdx++, i++)
       RefPicListTemp0[rIdx] = priv->RefPicSetLtCurr[i];
+    if (pps->pps_scc_extension_params.pps_curr_pic_ref_enabled_flag)
+      RefPicListTemp0[rIdx++] = picture;
   }
 
   /* construct RefPicList0 (8-9) */
@@ -1710,6 +1727,10 @@ init_picture_refs (GstVaapiDecoderH265 * decoder,
         ref_pic_list_modification->ref_pic_list_modification_flag_l0 ?
         RefPicListTemp0[ref_pic_list_modification->list_entry_l0[rIdx]] :
         RefPicListTemp0[rIdx];
+  if (pps->pps_scc_extension_params.pps_curr_pic_ref_enabled_flag
+      && !ref_pic_list_modification->ref_pic_list_modification_flag_l0
+      && (NumRpsCurrTempList0 > num_ref_idx_l0_active_minus1 + 1))
+    priv->RefPicList0[rIdx++] = picture;
   priv->RefPicList0_count = rIdx;
 
   if (type == GST_H265_B_SLICE) {
@@ -1726,6 +1747,8 @@ init_picture_refs (GstVaapiDecoderH265 * decoder,
       for (i = 0; i < priv->NumPocLtCurr && rIdx < NumRpsCurrTempList1;
           rIdx++, i++)
         RefPicListTemp1[rIdx] = priv->RefPicSetLtCurr[i];
+      if (pps->pps_scc_extension_params.pps_curr_pic_ref_enabled_flag)
+        RefPicListTemp1[rIdx++] = picture;
     }
 
     /* construct RefPicList1 (8-10) */
@@ -1898,6 +1921,17 @@ fill_picture (GstVaapiDecoderH265 * decoder, GstVaapiPictureH265 * picture)
     if (n >= G_N_ELEMENTS (pic_param->ReferenceFrames))
       break;
   }
+  /* 7.4.3.3.3, the current decoded picture is marked as "used for
+     long-term reference", no matter TwoVersionsOfCurrDecPicFlag */
+  if (pps->pps_scc_extension_params.pps_curr_pic_ref_enabled_flag
+      && n < G_N_ELEMENTS (pic_param->ReferenceFrames) - 1) {
+    gst_vaapi_picture_h265_set_reference (picture,
+        GST_VAAPI_PICTURE_FLAG_LONG_TERM_REFERENCE);
+    vaapi_fill_picture (&pic_param->ReferenceFrames[n++], picture,
+        picture->structure);
+    gst_vaapi_picture_h265_set_reference (picture, 0);
+  }
+
   for (; n < G_N_ELEMENTS (pic_param->ReferenceFrames); n++)
     vaapi_init_picture (&pic_param->ReferenceFrames[n]);
 
@@ -2277,6 +2311,7 @@ decode_ref_pic_set (GstVaapiDecoderH265 * decoder,
   GstVaapiDecoderH265Private *const priv = &decoder->priv;
   GstH265SliceHdr *const slice_hdr = &pi->data.slice_hdr;
   GstH265SPS *const sps = get_sps (decoder);
+  GstH265PPS *const pps = get_pps (decoder);
   const gint32 MaxPicOrderCntLsb =
       1 << (sps->log2_max_pic_order_cnt_lsb_minus4 + 4);
 
@@ -2297,6 +2332,7 @@ decode_ref_pic_set (GstVaapiDecoderH265 * decoder,
     memset (priv->PocLtFoll, 0, sizeof (guint) * 16);
     priv->NumPocStCurrBefore = priv->NumPocStCurrAfter = priv->NumPocStFoll = 0;
     priv->NumPocLtCurr = priv->NumPocLtFoll = 0;
+    priv->NumPocTotalCurr = 0;
   } else {
     GstH265ShortTermRefPicSet *stRefPic = NULL;
     gint32 num_lt_pics, pocLt;
@@ -2346,6 +2382,8 @@ decode_ref_pic_set (GstVaapiDecoderH265 * decoder,
         numtotalcurr++;
     }
 
+    if (pps->pps_scc_extension_params.pps_curr_pic_ref_enabled_flag)
+      numtotalcurr++;
     priv->NumPocTotalCurr = numtotalcurr;
 
     /* The variable DeltaPocMsbCycleLt[i] is derived as follows: (7-38) */