avcodec/cbs_vp9: keep track of reference frames

Signed-off-by: James Almer <jamrial@gmail.com>

libavcodec/cbs_vp9.h

@@ -181,6 +181,13 @@ typedef struct VP9RawSuperframe {
     VP9RawSuperframeIndex index;
 } VP9RawSuperframe;
 
+typedef struct VP9ReferenceFrameState {
+    int frame_width;    // RefFrameWidth
+    int frame_height;   // RefFrameHeight
+    int subsampling_x;  // RefSubsamplingX
+    int subsampling_y;  // RefSubsamplingY
+    int bit_depth;      // RefBitDepth
+} VP9ReferenceFrameState;
 
 typedef struct CodedBitstreamVP9Context {
     // Frame dimensions in 8x8 mode info blocks.
@@ -190,6 +197,15 @@ typedef struct CodedBitstreamVP9Context {
     uint16_t sb64_cols;
     uint16_t sb64_rows;
 
+    int frame_width;
+    int frame_height;
+
+    uint8_t subsampling_x;
+    uint8_t subsampling_y;
+    int bit_depth;
+
+    VP9ReferenceFrameState ref[VP9_NUM_REF_FRAMES];
+
     // Write buffer.
     uint8_t *write_buffer;
     size_t write_buffer_size;

libavcodec/cbs_vp9_syntax_template.c

@@ -43,10 +43,14 @@ static int FUNC(frame_sync_code)(CodedBitstreamContext *ctx, RWContext *rw,
 static int FUNC(color_config)(CodedBitstreamContext *ctx, RWContext *rw,
                               VP9RawFrameHeader *current, int profile)
 {
+    CodedBitstreamVP9Context *vp9 = ctx->priv_data;
     int err;
 
-    if (profile >= 2)
+    if (profile >= 2) {
         f(1, ten_or_twelve_bit);
+        vp9->bit_depth = current->ten_or_twelve_bit ? 12 : 10;
+    } else
+        vp9->bit_depth = 8;
 
     f(3, color_space);
 
@@ -69,6 +73,9 @@ static int FUNC(color_config)(CodedBitstreamContext *ctx, RWContext *rw,
         }
     }
 
+    vp9->subsampling_x = current->subsampling_x;
+    vp9->subsampling_y = current->subsampling_y;
+
     return 0;
 }
 
@@ -81,8 +88,11 @@ static int FUNC(frame_size)(CodedBitstreamContext *ctx, RWContext *rw,
     f(16, frame_width_minus_1);
     f(16, frame_height_minus_1);
 
-    vp9->mi_cols = (current->frame_width_minus_1  + 8) >> 3;
-    vp9->mi_rows = (current->frame_height_minus_1 + 8) >> 3;
+    vp9->frame_width  = current->frame_width_minus_1  + 1;
+    vp9->frame_height = current->frame_height_minus_1 + 1;
+
+    vp9->mi_cols = (vp9->frame_width  + 7) >> 3;
+    vp9->mi_rows = (vp9->frame_height + 7) >> 3;
     vp9->sb64_cols = (vp9->mi_cols + 7) >> 3;
     vp9->sb64_rows = (vp9->mi_rows + 7) >> 3;
 
@@ -107,15 +117,33 @@ static int FUNC(render_size)(CodedBitstreamContext *ctx, RWContext *rw,
 static int FUNC(frame_size_with_refs)(CodedBitstreamContext *ctx, RWContext *rw,
                                       VP9RawFrameHeader *current)
 {
+    CodedBitstreamVP9Context *vp9 = ctx->priv_data;
     int err, i;
 
     for (i = 0; i < VP9_REFS_PER_FRAME; i++) {
         fs(1, found_ref[i], 1, i);
-        if (current->found_ref[i])
+        if (current->found_ref[i]) {
+            VP9ReferenceFrameState *ref =
+                &vp9->ref[current->ref_frame_idx[i]];
+
+            vp9->frame_width   = ref->frame_width;
+            vp9->frame_height  = ref->frame_height;
+
+            vp9->subsampling_x = ref->subsampling_x;
+            vp9->subsampling_y = ref->subsampling_y;
+            vp9->bit_depth     = ref->bit_depth;
+
             break;
+        }
     }
     if (i >= VP9_REFS_PER_FRAME)
         CHECK(FUNC(frame_size)(ctx, rw, current));
+    else {
+        vp9->mi_cols = (vp9->frame_width  + 7) >> 3;
+        vp9->mi_rows = (vp9->frame_height + 7) >> 3;
+        vp9->sb64_cols = (vp9->mi_cols + 7) >> 3;
+        vp9->sb64_rows = (vp9->mi_rows + 7) >> 3;
+    }
     CHECK(FUNC(render_size)(ctx, rw, current));
 
     return 0;
@@ -249,6 +277,7 @@ static int FUNC(tile_info)(CodedBitstreamContext *ctx, RWContext *rw,
 static int FUNC(uncompressed_header)(CodedBitstreamContext *ctx, RWContext *rw,
                                      VP9RawFrameHeader *current)
 {
+    CodedBitstreamVP9Context *vp9 = ctx->priv_data;
     int profile, i;
     int err;
 
@@ -301,6 +330,10 @@ static int FUNC(uncompressed_header)(CodedBitstreamContext *ctx, RWContext *rw,
                  infer(color_space,   1);
                  infer(subsampling_x, 1);
                  infer(subsampling_y, 1);
+                 vp9->bit_depth = 8;
+
+                 vp9->subsampling_x = current->subsampling_x;
+                 vp9->subsampling_y = current->subsampling_y;
              }
 
              f(8, refresh_frame_flags);
@@ -339,6 +372,25 @@ static int FUNC(uncompressed_header)(CodedBitstreamContext *ctx, RWContext *rw,
 
     f(16, header_size_in_bytes);
 
+    for (i = 0; i < VP9_NUM_REF_FRAMES; i++) {
+        if (current->refresh_frame_flags & (1 << i)) {
+            vp9->ref[i] = (VP9ReferenceFrameState) {
+                .frame_width    = vp9->frame_width,
+                .frame_height   = vp9->frame_height,
+                .subsampling_x  = vp9->subsampling_x,
+                .subsampling_y  = vp9->subsampling_y,
+                .bit_depth      = vp9->bit_depth,
+            };
+        }
+    }
+
+    av_log(ctx->log_ctx, AV_LOG_DEBUG, "Frame:  size %dx%d  "
+           "subsample %dx%d  bit_depth %d  tiles %dx%d.\n",
+           vp9->frame_width, vp9->frame_height,
+           vp9->subsampling_x, vp9->subsampling_y,
+           vp9->bit_depth, 1 << current->tile_cols_log2,
+           1 << current->tile_rows_log2);
+
     return 0;
 }