+static void vc1_loop_filter_iblk_delayed(VC1Context *v, int pq)
+{
+ MpegEncContext *s = &v->s;
+ int j;
+
+ /* The loopfilter runs 1 row and 1 column behind the overlap filter, which
+ * means it runs two rows/cols behind the decoding loop. */
+ if (!s->first_slice_line) {
+ if (s->mb_x) {
+ if (s->mb_y >= s->start_mb_y + 2) {
+ v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);
+
+ if (s->mb_x >= 2)
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 16, s->linesize, pq);
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize - 8, s->linesize, pq);
+ for(j = 0; j < 2; j++) {
+ v->vc1dsp.vc1_v_loop_filter8(s->dest[j+1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);
+ if (s->mb_x >= 2) {
+ v->vc1dsp.vc1_h_loop_filter8(s->dest[j+1] - 16 * s->uvlinesize - 8, s->uvlinesize, pq);
+ }
+ }
+ }
+ v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize - 16, s->linesize, pq);
+ }
+
+ if (s->mb_x == s->mb_width - 1) {
+ if (s->mb_y >= s->start_mb_y + 2) {
+ v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
+
+ if (s->mb_x)
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize, s->linesize, pq);
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 32 * s->linesize + 8, s->linesize, pq);
+ for(j = 0; j < 2; j++) {
+ v->vc1dsp.vc1_v_loop_filter8(s->dest[j+1] - 8 * s->uvlinesize, s->uvlinesize, pq);
+ if (s->mb_x >= 2) {
+ v->vc1dsp.vc1_h_loop_filter8(s->dest[j+1] - 16 * s->uvlinesize, s->uvlinesize, pq);
+ }
+ }
+ }
+ v->vc1dsp.vc1_v_loop_filter16(s->dest[0] - 8 * s->linesize, s->linesize, pq);
+ }
+
+ if (s->mb_y == s->mb_height) {
+ if (s->mb_x) {
+ if (s->mb_x >= 2)
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 16, s->linesize, pq);
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 8, s->linesize, pq);
+ if (s->mb_x >= 2) {
+ for(j = 0; j < 2; j++) {
+ v->vc1dsp.vc1_h_loop_filter8(s->dest[j+1] - 8 * s->uvlinesize - 8, s->uvlinesize, pq);
+ }
+ }
+ }
+
+ if (s->mb_x == s->mb_width - 1) {
+ if (s->mb_x)
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize, s->linesize, pq);
+ v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize + 8, s->linesize, pq);
+ if (s->mb_x) {
+ for(j = 0; j < 2; j++) {
+ v->vc1dsp.vc1_h_loop_filter8(s->dest[j+1] - 8 * s->uvlinesize, s->uvlinesize, pq);
+ }
+ }
+ }
+ }
+ }
+}
+
+static void vc1_smooth_overlap_filter_iblk(VC1Context *v)
+{
+ MpegEncContext *s = &v->s;
+ int mb_pos;
+
+ if (v->condover == CONDOVER_NONE)
+ return;
+
+ mb_pos = s->mb_x + s->mb_y * s->mb_stride;
+
+ /* Within a MB, the horizontal overlap always runs before the vertical.
+ * To accomplish that, we run the H on left and internal borders of the
+ * currently decoded MB. Then, we wait for the next overlap iteration
+ * to do H overlap on the right edge of this MB, before moving over and
+ * running the V overlap. Therefore, the V overlap makes us trail by one
+ * MB col and the H overlap filter makes us trail by one MB row. This
+ * is reflected in the time at which we run the put_pixels loop. */
+ if(v->condover == CONDOVER_ALL || v->pq >= 9 || v->over_flags_plane[mb_pos]) {
+ if(s->mb_x && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
+ v->over_flags_plane[mb_pos - 1])) {
+ v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][1],
+ v->block[v->cur_blk_idx][0]);
+ v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][3],
+ v->block[v->cur_blk_idx][2]);
+ if(!(s->flags & CODEC_FLAG_GRAY)) {
+ v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][4],
+ v->block[v->cur_blk_idx][4]);
+ v->vc1dsp.vc1_h_s_overlap(v->block[v->left_blk_idx][5],
+ v->block[v->cur_blk_idx][5]);
+ }
+ }
+ v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][0],
+ v->block[v->cur_blk_idx][1]);
+ v->vc1dsp.vc1_h_s_overlap(v->block[v->cur_blk_idx][2],
+ v->block[v->cur_blk_idx][3]);
+
+ if (s->mb_x == s->mb_width - 1) {
+ if(!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
+ v->over_flags_plane[mb_pos - s->mb_stride])) {
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][2],
+ v->block[v->cur_blk_idx][0]);
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][3],
+ v->block[v->cur_blk_idx][1]);
+ if(!(s->flags & CODEC_FLAG_GRAY)) {
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][4],
+ v->block[v->cur_blk_idx][4]);
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->top_blk_idx][5],
+ v->block[v->cur_blk_idx][5]);
+ }
+ }
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][0],
+ v->block[v->cur_blk_idx][2]);
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->cur_blk_idx][1],
+ v->block[v->cur_blk_idx][3]);
+ }
+ }
+ if (s->mb_x && (v->condover == CONDOVER_ALL || v->over_flags_plane[mb_pos - 1])) {
+ if(!s->first_slice_line && (v->condover == CONDOVER_ALL || v->pq >= 9 ||
+ v->over_flags_plane[mb_pos - s->mb_stride - 1])) {
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][2],
+ v->block[v->left_blk_idx][0]);
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][3],
+ v->block[v->left_blk_idx][1]);
+ if(!(s->flags & CODEC_FLAG_GRAY)) {
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][4],
+ v->block[v->left_blk_idx][4]);
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->topleft_blk_idx][5],
+ v->block[v->left_blk_idx][5]);
+ }
+ }
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][0],
+ v->block[v->left_blk_idx][2]);
+ v->vc1dsp.vc1_v_s_overlap(v->block[v->left_blk_idx][1],
+ v->block[v->left_blk_idx][3]);
+ }
+}
+