vp8: refactor decoding a single mb_row

This is in preperation for sliced threading.
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

libavcodec/vp8.c

@@ -1574,11 +1574,95 @@ static void release_queued_segmaps(VP8Context *s, int is_close)
     s->maps_are_invalid = 0;
 }
 
+#define MARGIN (16 << 2)
+static void vp8_decode_mb_row(AVCodecContext *avctx, AVFrame *curframe,
+                              AVFrame *prev_frame, int mb_y)
+{
+    VP8Context *s = avctx->priv_data;
+    VP56RangeCoder *c = &s->coeff_partition[mb_y & (s->num_coeff_partitions-1)];
+    VP8Macroblock *mb = s->macroblocks + (s->mb_height - mb_y - 1)*2;
+    int i, y, mb_x, mb_xy = mb_y*s->mb_width;
+    uint8_t *dst[3] = {
+        curframe->data[0] + 16*mb_y*s->linesize,
+        curframe->data[1] +  8*mb_y*s->uvlinesize,
+        curframe->data[2] +  8*mb_y*s->uvlinesize
+    };
+
+    memset(mb - 1, 0, sizeof(*mb));   // zero left macroblock
+    memset(s->left_nnz, 0, sizeof(s->left_nnz));
+    AV_WN32A(s->intra4x4_pred_mode_left, DC_PRED*0x01010101);
+
+    // left edge of 129 for intra prediction
+    if (!(avctx->flags & CODEC_FLAG_EMU_EDGE)) {
+        for (i = 0; i < 3; i++)
+            for (y = 0; y < 16>>!!i; y++)
+                dst[i][y*curframe->linesize[i]-1] = 129;
+        if (mb_y == 1) // top left edge is also 129
+            s->top_border[0][15] = s->top_border[0][23] = s->top_border[0][31] = 129;
+    }
+
+    s->mv_min.x = -MARGIN;
+    s->mv_max.x = ((s->mb_width  - 1) << 6) + MARGIN;
+
+    for (mb_x = 0; mb_x < s->mb_width; mb_x++, mb_xy++, mb++) {
+        /* Prefetch the current frame, 4 MBs ahead */
+        s->dsp.prefetch(dst[0] + (mb_x&3)*4*s->linesize + 64, s->linesize, 4);
+        s->dsp.prefetch(dst[1] + (mb_x&7)*s->uvlinesize + 64, dst[2] - dst[1], 2);
+
+        decode_mb_mode(s, mb, mb_x, mb_y, curframe->ref_index[0] + mb_xy,
+                       prev_frame && prev_frame->ref_index[0] ? prev_frame->ref_index[0] + mb_xy : NULL);
+
+        prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_PREVIOUS);
+
+        if (!mb->skip)
+            decode_mb_coeffs(s, c, mb, s->top_nnz[mb_x], s->left_nnz);
+
+        if (mb->mode <= MODE_I4x4)
+            intra_predict(s, dst, mb, mb_x, mb_y);
+        else
+            inter_predict(s, dst, mb, mb_x, mb_y);
+
+        prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_GOLDEN);
+
+        if (!mb->skip) {
+            idct_mb(s, dst, mb);
+        } else {
+            AV_ZERO64(s->left_nnz);
+            AV_WN64(s->top_nnz[mb_x], 0);   // array of 9, so unaligned
+
+            // Reset DC block predictors if they would exist if the mb had coefficients
+            if (mb->mode != MODE_I4x4 && mb->mode != VP8_MVMODE_SPLIT) {
+                s->left_nnz[8]      = 0;
+                s->top_nnz[mb_x][8] = 0;
+            }
+        }
+
+        if (s->deblock_filter)
+            filter_level_for_mb(s, mb, &s->filter_strength[mb_x]);
+
+        prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_GOLDEN2);
+
+        dst[0] += 16;
+        dst[1] += 8;
+        dst[2] += 8;
+        s->mv_min.x -= 64;
+        s->mv_max.x -= 64;
+    }
+    if (s->deblock_filter) {
+        if (s->filter.simple)
+            filter_mb_row_simple(s, curframe, mb_y);
+        else
+            filter_mb_row(s, curframe, mb_y);
+    }
+    s->mv_min.y -= 64;
+    s->mv_max.y -= 64;
+}
+
 static int vp8_decode_frame(AVCodecContext *avctx, void *data, int *data_size,
                             AVPacket *avpkt)
 {
     VP8Context *s = avctx->priv_data;
-    int ret, mb_x, mb_y, i, y, referenced;
+    int ret, mb_y, i, referenced;
     enum AVDiscard skip_thresh;
     AVFrame *av_uninit(curframe), *prev_frame;
 
@@ -1686,90 +1770,14 @@ static int vp8_decode_frame(AVCodecContext *avctx, void *data, int *data_size,
     if (s->keyframe)
         memset(s->intra4x4_pred_mode_top, DC_PRED, s->mb_width*4);
 
-#define MARGIN (16 << 2)
     s->mv_min.y = -MARGIN;
     s->mv_max.y = ((s->mb_height - 1) << 6) + MARGIN;
 
     for (mb_y = 0; mb_y < s->mb_height; mb_y++) {
-        VP56RangeCoder *c = &s->coeff_partition[mb_y & (s->num_coeff_partitions-1)];
-        VP8Macroblock *mb = s->macroblocks + (s->mb_height - mb_y - 1)*2;
-        int mb_xy = mb_y*s->mb_width;
-        uint8_t *dst[3] = {
-            curframe->data[0] + 16*mb_y*s->linesize,
-            curframe->data[1] +  8*mb_y*s->uvlinesize,
-            curframe->data[2] +  8*mb_y*s->uvlinesize
-        };
-
-        memset(mb - 1, 0, sizeof(*mb));   // zero left macroblock
-        memset(s->left_nnz, 0, sizeof(s->left_nnz));
-        AV_WN32A(s->intra4x4_pred_mode_left, DC_PRED*0x01010101);
-
-        // left edge of 129 for intra prediction
-        if (!(avctx->flags & CODEC_FLAG_EMU_EDGE)) {
-            for (i = 0; i < 3; i++)
-                for (y = 0; y < 16>>!!i; y++)
-                    dst[i][y*curframe->linesize[i]-1] = 129;
-            if (mb_y == 1) // top left edge is also 129
-                s->top_border[0][15] = s->top_border[0][23] = s->top_border[0][31] = 129;
-        }
-
-        s->mv_min.x = -MARGIN;
-        s->mv_max.x = ((s->mb_width  - 1) << 6) + MARGIN;
         if (prev_frame && s->segmentation.enabled && !s->segmentation.update_map)
             ff_thread_await_progress(prev_frame, mb_y, 0);
 
-        for (mb_x = 0; mb_x < s->mb_width; mb_x++, mb_xy++, mb++) {
-            /* Prefetch the current frame, 4 MBs ahead */
-            s->dsp.prefetch(dst[0] + (mb_x&3)*4*s->linesize + 64, s->linesize, 4);
-            s->dsp.prefetch(dst[1] + (mb_x&7)*s->uvlinesize + 64, dst[2] - dst[1], 2);
-
-            decode_mb_mode(s, mb, mb_x, mb_y, curframe->ref_index[0] + mb_xy,
-                           prev_frame && prev_frame->ref_index[0] ? prev_frame->ref_index[0] + mb_xy : NULL);
-
-            prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_PREVIOUS);
-
-            if (!mb->skip)
-                decode_mb_coeffs(s, c, mb, s->top_nnz[mb_x], s->left_nnz);
-
-            if (mb->mode <= MODE_I4x4)
-                intra_predict(s, dst, mb, mb_x, mb_y);
-            else
-                inter_predict(s, dst, mb, mb_x, mb_y);
-
-            prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_GOLDEN);
-
-            if (!mb->skip) {
-                idct_mb(s, dst, mb);
-            } else {
-                AV_ZERO64(s->left_nnz);
-                AV_WN64(s->top_nnz[mb_x], 0);   // array of 9, so unaligned
-
-                // Reset DC block predictors if they would exist if the mb had coefficients
-                if (mb->mode != MODE_I4x4 && mb->mode != VP8_MVMODE_SPLIT) {
-                    s->left_nnz[8]      = 0;
-                    s->top_nnz[mb_x][8] = 0;
-                }
-            }
-
-            if (s->deblock_filter)
-                filter_level_for_mb(s, mb, &s->filter_strength[mb_x]);
-
-            prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_GOLDEN2);
-
-            dst[0] += 16;
-            dst[1] += 8;
-            dst[2] += 8;
-            s->mv_min.x -= 64;
-            s->mv_max.x -= 64;
-        }
-        if (s->deblock_filter) {
-            if (s->filter.simple)
-                filter_mb_row_simple(s, curframe, mb_y);
-            else
-                filter_mb_row(s, curframe, mb_y);
-        }
-        s->mv_min.y -= 64;
-        s->mv_max.y -= 64;
+        vp8_decode_mb_row(avctx, curframe, prev_frame, mb_y);
 
         ff_thread_report_progress(curframe, mb_y, 0);
     }