vp9: fix scaled motion vector clipping for sub8x8 blocks.

To match the obscure clipping bug behaviour in libvpx.

2 files changed, 70 insertions, 62 deletions

diff --git a/libavcodec/vp9.c b/libavcodec/vp9.c
index 42c1ec97e3..09c96f162a 100644
--- a/libavcodec/vp9.c
+++ b/libavcodec/vp9.c
@@ -2750,6 +2750,7 @@ static av_always_inline void mc_luma_scaled(VP9Context *s, vp9_scaled_mc_func sm
                                             const uint8_t *ref, ptrdiff_t ref_stride,
                                             ThreadFrame *ref_frame,
                                             ptrdiff_t y, ptrdiff_t x, const VP56mv *in_mv,
+                                            int px, int py, int pw, int ph,
                                             int bw, int bh, int w, int h, int bytesperpixel,
                                             const uint16_t *scale, const uint8_t *step)
 {
@@ -2759,8 +2760,8 @@ static av_always_inline void mc_luma_scaled(VP9Context *s, vp9_scaled_mc_func sm
     int th;
     VP56mv mv;
 
-    mv.x = av_clip(in_mv->x, -(x + bw + 4) << 3, (s->cols * 8 - x + 3) << 3);
-    mv.y = av_clip(in_mv->y, -(y + bh + 4) << 3, (s->rows * 8 - y + 3) << 3);
+    mv.x = av_clip(in_mv->x, -(x + pw - px + 4) << 3, (s->cols * 8 - x + px + 3) << 3);
+    mv.y = av_clip(in_mv->y, -(y + ph - py + 4) << 3, (s->rows * 8 - y + py + 3) << 3);
     // BUG libvpx seems to scale the two components separately. This introduces
     // rounding errors but we have to reproduce them to be exactly compatible
     // with the output from libvpx...
@@ -2798,6 +2799,7 @@ static av_always_inline void mc_chroma_scaled(VP9Context *s, vp9_scaled_mc_func
                                               const uint8_t *ref_v, ptrdiff_t src_stride_v,
                                               ThreadFrame *ref_frame,
                                               ptrdiff_t y, ptrdiff_t x, const VP56mv *in_mv,
+                                              int px, int py, int pw, int ph,
                                               int bw, int bh, int w, int h, int bytesperpixel,
                                               const uint16_t *scale, const uint8_t *step)
 {
@@ -2808,18 +2810,18 @@ static av_always_inline void mc_chroma_scaled(VP9Context *s, vp9_scaled_mc_func
 
     if (s->ss_h) {
         // BUG https://code.google.com/p/webm/issues/detail?id=820
-        mv.x = av_clip(in_mv->x, -(x + bw + 4) << 4, (s->cols * 4 - x + 3) << 4);
+        mv.x = av_clip(in_mv->x, -(x + pw - px + 4) << 4, (s->cols * 4 - x + px + 3) << 4);
         mx = scale_mv(mv.x, 0) + (scale_mv(x * 16, 0) & ~15) + (scale_mv(x * 32, 0) & 15);
     } else {
-        mv.x = av_clip(in_mv->x, -(x + bw + 4) << 3, (s->cols * 8 - x + 3) << 3);
+        mv.x = av_clip(in_mv->x, -(x + pw - px + 4) << 3, (s->cols * 8 - x + px + 3) << 3);
         mx = scale_mv(mv.x << 1, 0) + scale_mv(x * 16, 0);
     }
     if (s->ss_v) {
         // BUG https://code.google.com/p/webm/issues/detail?id=820
-        mv.y = av_clip(in_mv->y, -(y + bh + 4) << 4, (s->rows * 4 - y + 3) << 4);
+        mv.y = av_clip(in_mv->y, -(y + ph - py + 4) << 4, (s->rows * 4 - y + py + 3) << 4);
         my = scale_mv(mv.y, 1) + (scale_mv(y * 16, 1) & ~15) + (scale_mv(y * 32, 1) & 15);
     } else {
-        mv.y = av_clip(in_mv->y, -(y + bh + 4) << 3, (s->rows * 8 - y + 3) << 3);
+        mv.y = av_clip(in_mv->y, -(y + ph - py + 4) << 3, (s->rows * 8 - y + py + 3) << 3);
         my = scale_mv(mv.y << 1, 1) + scale_mv(y * 16, 1);
     }
 #undef scale_mv
@@ -2858,14 +2860,15 @@ static av_always_inline void mc_chroma_scaled(VP9Context *s, vp9_scaled_mc_func
     }
 }
 
-#define mc_luma_dir(s, mc, dst, dst_ls, src, src_ls, tref, row, col, mv, bw, bh, w, h, i) \
+#define mc_luma_dir(s, mc, dst, dst_ls, src, src_ls, tref, row, col, mv, \
+                    px, py, pw, ph, bw, bh, w, h, i) \
     mc_luma_scaled(s, s->dsp.s##mc, dst, dst_ls, src, src_ls, tref, row, col, \
-                   mv, bw, bh, w, h, bytesperpixel, \
+                   mv, px, py, pw, ph, bw, bh, w, h, bytesperpixel, \
                    s->mvscale[b->ref[i]], s->mvstep[b->ref[i]])
 #define mc_chroma_dir(s, mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
-                      row, col, mv, bw, bh, w, h, i) \
+                      row, col, mv, px, py, pw, ph, bw, bh, w, h, i) \
     mc_chroma_scaled(s, s->dsp.s##mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
-                     row, col, mv, bw, bh, w, h, bytesperpixel, \
+                     row, col, mv, px, py, pw, ph, bw, bh, w, h, bytesperpixel, \
                      s->mvscale[b->ref[i]], s->mvstep[b->ref[i]])
 #define SCALED 1
 #define FN(x) x##_scaled_8bpp
@@ -2959,11 +2962,12 @@ static av_always_inline void mc_chroma_unscaled(VP9Context *s, vp9_mc_func (*mc)
     }
 }
 
-#define mc_luma_dir(s, mc, dst, dst_ls, src, src_ls, tref, row, col, mv, bw, bh, w, h, i) \
+#define mc_luma_dir(s, mc, dst, dst_ls, src, src_ls, tref, row, col, mv, \
+                    px, py, pw, ph, bw, bh, w, h, i) \
     mc_luma_unscaled(s, s->dsp.mc, dst, dst_ls, src, src_ls, tref, row, col, \
                      mv, bw, bh, w, h, bytesperpixel)
 #define mc_chroma_dir(s, mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
-                      row, col, mv, bw, bh, w, h, i) \
+                      row, col, mv, px, py, pw, ph, bw, bh, w, h, i) \
     mc_chroma_unscaled(s, s->dsp.mc, dstu, dstv, dst_ls, srcu, srcu_ls, srcv, srcv_ls, tref, \
                        row, col, mv, bw, bh, w, h, bytesperpixel)
 #define SCALED 0
diff --git a/libavcodec/vp9_mc_template.c b/libavcodec/vp9_mc_template.c
index 5e6ee870b8..f4eb4e56ac 100644
--- a/libavcodec/vp9_mc_template.c
+++ b/libavcodec/vp9_mc_template.c
@@ -57,11 +57,11 @@ static void FN(inter_pred)(AVCodecContext *ctx)
         if (b->bs == BS_8x4) {
             mc_luma_dir(s, mc[3][b->filter][0], s->dst[0], ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        row << 3, col << 3, &b->mv[0][0], 8, 4, w1, h1, 0);
+                        row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);
             mc_luma_dir(s, mc[3][b->filter][0],
                         s->dst[0] + 4 * ls_y, ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        (row << 3) + 4, col << 3, &b->mv[2][0], 8, 4, w1, h1, 0);
+                        (row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);
             w1 = (w1 + s->ss_h) >> s->ss_h;
             if (s->ss_v) {
                 h1 = (h1 + 1) >> 1;
@@ -71,14 +71,14 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                               ref1->data[1], ref1->linesize[1],
                               ref1->data[2], ref1->linesize[2], tref1,
                               row << 2, col << (3 - s->ss_h),
-                              &uvmv, 8 >> s->ss_h, 4, w1, h1, 0);
+                              &uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
             } else {
                 mc_chroma_dir(s, mc[3 + s->ss_h][b->filter][0],
                               s->dst[1], s->dst[2], ls_uv,
                               ref1->data[1], ref1->linesize[1],
                               ref1->data[2], ref1->linesize[2], tref1,
                               row << 3, col << (3 - s->ss_h),
-                              &b->mv[0][0], 8 >> s->ss_h, 4, w1, h1, 0);
+                              &b->mv[0][0],,,,, 8 >> s->ss_h, 4, w1, h1, 0);
                 // BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
                 // to get the motion vector for the bottom 4x4 block
                 // https://code.google.com/p/webm/issues/detail?id=993
@@ -92,17 +92,17 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                               ref1->data[1], ref1->linesize[1],
                               ref1->data[2], ref1->linesize[2], tref1,
                               (row << 3) + 4, col << (3 - s->ss_h),
-                              &uvmv, 8 >> s->ss_h, 4, w1, h1, 0);
+                              &uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
             }
 
             if (b->comp) {
                 mc_luma_dir(s, mc[3][b->filter][1], s->dst[0], ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            row << 3, col << 3, &b->mv[0][1], 8, 4, w2, h2, 1);
+                            row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);
                 mc_luma_dir(s, mc[3][b->filter][1],
                             s->dst[0] + 4 * ls_y, ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            (row << 3) + 4, col << 3, &b->mv[2][1], 8, 4, w2, h2, 1);
+                            (row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);
                 w2 = (w2 + s->ss_h) >> s->ss_h;
                 if (s->ss_v) {
                     h2 = (h2 + 1) >> 1;
@@ -112,14 +112,14 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref2->data[1], ref2->linesize[1],
                                   ref2->data[2], ref2->linesize[2], tref2,
                                   row << 2, col << (3 - s->ss_h),
-                                  &uvmv, 8 >> s->ss_h, 4, w2, h2, 1);
+                                  &uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
                 } else {
                     mc_chroma_dir(s, mc[3 + s->ss_h][b->filter][1],
                                   s->dst[1], s->dst[2], ls_uv,
                                   ref2->data[1], ref2->linesize[1],
                                   ref2->data[2], ref2->linesize[2], tref2,
                                   row << 3, col << (3 - s->ss_h),
-                                  &b->mv[0][1], 8 >> s->ss_h, 4, w2, h2, 1);
+                                  &b->mv[0][1],,,,, 8 >> s->ss_h, 4, w2, h2, 1);
                     // BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
                     // to get the motion vector for the bottom 4x4 block
                     // https://code.google.com/p/webm/issues/detail?id=993
@@ -133,16 +133,16 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref2->data[1], ref2->linesize[1],
                                   ref2->data[2], ref2->linesize[2], tref2,
                                   (row << 3) + 4, col << (3 - s->ss_h),
-                                  &uvmv, 8 >> s->ss_h, 4, w2, h2, 1);
+                                  &uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
                 }
             }
         } else if (b->bs == BS_4x8) {
             mc_luma_dir(s, mc[4][b->filter][0], s->dst[0], ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        row << 3, col << 3, &b->mv[0][0], 4, 8, w1, h1, 0);
+                        row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);
             mc_luma_dir(s, mc[4][b->filter][0], s->dst[0] + 4 * bytesperpixel, ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        row << 3, (col << 3) + 4, &b->mv[1][0], 4, 8, w1, h1, 0);
+                        row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);
             h1 = (h1 + s->ss_v) >> s->ss_v;
             if (s->ss_h) {
                 w1 = (w1 + 1) >> 1;
@@ -152,30 +152,30 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                               ref1->data[1], ref1->linesize[1],
                               ref1->data[2], ref1->linesize[2], tref1,
                               row << (3 - s->ss_v), col << 2,
-                              &uvmv, 4, 8 >> s->ss_v, w1, h1, 0);
+                              &uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);
             } else {
                 mc_chroma_dir(s, mc[4][b->filter][0],
                               s->dst[1], s->dst[2], ls_uv,
                               ref1->data[1], ref1->linesize[1],
                               ref1->data[2], ref1->linesize[2], tref1,
                               row << (3 - s->ss_v), col << 3,
-                              &b->mv[0][0], 4, 8 >> s->ss_v, w1, h1, 0);
+                              &b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
                 mc_chroma_dir(s, mc[4][b->filter][0],
                               s->dst[1] + 4 * bytesperpixel,
                               s->dst[2] + 4 * bytesperpixel, ls_uv,
                               ref1->data[1], ref1->linesize[1],
                               ref1->data[2], ref1->linesize[2], tref1,
                               row << (3 - s->ss_v), (col << 3) + 4,
-                              &b->mv[1][0], 4, 8 >> s->ss_v, w1, h1, 0);
+                              &b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
             }
 
             if (b->comp) {
                 mc_luma_dir(s, mc[4][b->filter][1], s->dst[0], ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            row << 3, col << 3, &b->mv[0][1], 4, 8, w2, h2, 1);
+                            row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);
                 mc_luma_dir(s, mc[4][b->filter][1], s->dst[0] + 4 * bytesperpixel, ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            row << 3, (col << 3) + 4, &b->mv[1][1], 4, 8, w2, h2, 1);
+                            row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);
                 h2 = (h2 + s->ss_v) >> s->ss_v;
                 if (s->ss_h) {
                     w2 = (w2 + 1) >> 1;
@@ -185,21 +185,21 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref2->data[1], ref2->linesize[1],
                                   ref2->data[2], ref2->linesize[2], tref2,
                                   row << (3 - s->ss_v), col << 2,
-                                  &uvmv, 4, 8 >> s->ss_v, w2, h2, 1);
+                                  &uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);
                 } else {
                     mc_chroma_dir(s, mc[4][b->filter][1],
                                   s->dst[1], s->dst[2], ls_uv,
                                   ref2->data[1], ref2->linesize[1],
                                   ref2->data[2], ref2->linesize[2], tref2,
                                   row << (3 - s->ss_v), col << 3,
-                                  &b->mv[0][1], 4, 8 >> s->ss_v, w2, h2, 1);
+                                  &b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
                     mc_chroma_dir(s, mc[4][b->filter][1],
                                   s->dst[1] + 4 * bytesperpixel,
                                   s->dst[2] + 4 * bytesperpixel, ls_uv,
                                   ref2->data[1], ref2->linesize[1],
                                   ref2->data[2], ref2->linesize[2], tref2,
                                   row << (3 - s->ss_v), (col << 3) + 4,
-                                  &b->mv[1][1], 4, 8 >> s->ss_v, w2, h2, 1);
+                                  &b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
                 }
             }
         } else
@@ -211,18 +211,22 @@ static void FN(inter_pred)(AVCodecContext *ctx)
             // do a w8 instead of a w4 call
             mc_luma_dir(s, mc[4][b->filter][0], s->dst[0], ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        row << 3, col << 3, &b->mv[0][0], 4, 4, w1, h1, 0);
+                        row << 3, col << 3, &b->mv[0][0],
+                        0, 0, 8, 8, 4, 4, w1, h1, 0);
             mc_luma_dir(s, mc[4][b->filter][0], s->dst[0] + 4 * bytesperpixel, ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        row << 3, (col << 3) + 4, &b->mv[1][0], 4, 4, w1, h1, 0);
+                        row << 3, (col << 3) + 4, &b->mv[1][0],
+                        4, 0, 8, 8, 4, 4, w1, h1, 0);
             mc_luma_dir(s, mc[4][b->filter][0],
                         s->dst[0] + 4 * ls_y, ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        (row << 3) + 4, col << 3, &b->mv[2][0], 4, 4, w1, h1, 0);
+                        (row << 3) + 4, col << 3, &b->mv[2][0],
+                        0, 4, 8, 8, 4, 4, w1, h1, 0);
             mc_luma_dir(s, mc[4][b->filter][0],
                         s->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
                         ref1->data[0], ref1->linesize[0], tref1,
-                        (row << 3) + 4, (col << 3) + 4, &b->mv[3][0], 4, 4, w1, h1, 0);
+                        (row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
+                        4, 4, 8, 8, 4, 4, w1, h1, 0);
             if (s->ss_v) {
                 h1 = (h1 + 1) >> 1;
                 if (s->ss_h) {
@@ -234,7 +238,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   row << 2, col << 2,
-                                  &uvmv, 4, 4, w1, h1, 0);
+                                  &uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
                 } else {
                     uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
                     mc_chroma_dir(s, mc[4][b->filter][0],
@@ -242,7 +246,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   row << 2, col << 3,
-                                  &uvmv, 4, 4, w1, h1, 0);
+                                  &uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
                     uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
                     mc_chroma_dir(s, mc[4][b->filter][0],
                                   s->dst[1] + 4 * bytesperpixel,
@@ -250,7 +254,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   row << 2, (col << 3) + 4,
-                                  &uvmv, 4, 4, w1, h1, 0);
+                                  &uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
                 }
             } else {
                 if (s->ss_h) {
@@ -261,7 +265,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   row << 3, col << 2,
-                                  &uvmv, 4, 4, w1, h1, 0);
+                                  &uvmv, 0, 0, 4, 8, 4, 4, w1, h1, 0);
                     // BUG libvpx uses wrong block index for 4:2:2 bs=4x4
                     // bottom block
                     // https://code.google.com/p/webm/issues/detail?id=993
@@ -271,52 +275,52 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   (row << 3) + 4, col << 2,
-                                  &uvmv, 4, 4, w1, h1, 0);
+                                  &uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
                 } else {
                     mc_chroma_dir(s, mc[4][b->filter][0],
                                   s->dst[1], s->dst[2], ls_uv,
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   row << 3, col << 3,
-                                  &b->mv[0][0], 4, 4, w1, h1, 0);
+                                  &b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
                     mc_chroma_dir(s, mc[4][b->filter][0],
                                   s->dst[1] + 4 * bytesperpixel,
                                   s->dst[2] + 4 * bytesperpixel, ls_uv,
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   row << 3, (col << 3) + 4,
-                                  &b->mv[1][0], 4, 4, w1, h1, 0);
+                                  &b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
                     mc_chroma_dir(s, mc[4][b->filter][0],
                                   s->dst[1] + 4 * ls_uv, s->dst[2] + 4 * ls_uv, ls_uv,
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   (row << 3) + 4, col << 3,
-                                  &b->mv[2][0], 4, 4, w1, h1, 0);
+                                  &b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
                     mc_chroma_dir(s, mc[4][b->filter][0],
                                   s->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
                                   s->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
                                   ref1->data[1], ref1->linesize[1],
                                   ref1->data[2], ref1->linesize[2], tref1,
                                   (row << 3) + 4, (col << 3) + 4,
-                                  &b->mv[3][0], 4, 4, w1, h1, 0);
+                                  &b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
                 }
             }
 
             if (b->comp) {
                 mc_luma_dir(s, mc[4][b->filter][1], s->dst[0], ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            row << 3, col << 3, &b->mv[0][1], 4, 4, w2, h2, 1);
+                            row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
                 mc_luma_dir(s, mc[4][b->filter][1], s->dst[0] + 4 * bytesperpixel, ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            row << 3, (col << 3) + 4, &b->mv[1][1], 4, 4, w2, h2, 1);
+                            row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
                 mc_luma_dir(s, mc[4][b->filter][1],
                             s->dst[0] + 4 * ls_y, ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            (row << 3) + 4, col << 3, &b->mv[2][1], 4, 4, w2, h2, 1);
+                            (row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
                 mc_luma_dir(s, mc[4][b->filter][1],
                             s->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
                             ref2->data[0], ref2->linesize[0], tref2,
-                            (row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, w2, h2, 1);
+                            (row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
                 if (s->ss_v) {
                     h2 = (h2 + 1) >> 1;
                     if (s->ss_h) {
@@ -328,7 +332,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       row << 2, col << 2,
-                                      &uvmv, 4, 4, w2, h2, 1);
+                                      &uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
                     } else {
                         uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
                         mc_chroma_dir(s, mc[4][b->filter][1],
@@ -336,7 +340,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       row << 2, col << 3,
-                                      &uvmv, 4, 4, w2, h2, 1);
+                                      &uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
                         uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
                         mc_chroma_dir(s, mc[4][b->filter][1],
                                       s->dst[1] + 4 * bytesperpixel,
@@ -344,7 +348,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       row << 2, (col << 3) + 4,
-                                      &uvmv, 4, 4, w2, h2, 1);
+                                      &uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
                     }
                 } else {
                     if (s->ss_h) {
@@ -355,7 +359,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       row << 3, col << 2,
-                                      &uvmv, 4, 4, w2, h2, 1);
+                                      &uvmv, 0, 0, 4, 8, 4, 4, w2, h2, 1);
                         // BUG libvpx uses wrong block index for 4:2:2 bs=4x4
                         // bottom block
                         // https://code.google.com/p/webm/issues/detail?id=993
@@ -365,34 +369,34 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       (row << 3) + 4, col << 2,
-                                      &uvmv, 4, 4, w2, h2, 1);
+                                      &uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
                     } else {
                         mc_chroma_dir(s, mc[4][b->filter][1],
                                       s->dst[1], s->dst[2], ls_uv,
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       row << 3, col << 3,
-                                      &b->mv[0][1], 4, 4, w2, h2, 1);
+                                      &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
                         mc_chroma_dir(s, mc[4][b->filter][1],
                                       s->dst[1] + 4 * bytesperpixel,
                                       s->dst[2] + 4 * bytesperpixel, ls_uv,
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       row << 3, (col << 3) + 4,
-                                      &b->mv[1][1], 4, 4, w2, h2, 1);
+                                      &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
                         mc_chroma_dir(s, mc[4][b->filter][1],
                                       s->dst[1] + 4 * ls_uv, s->dst[2] + 4 * ls_uv, ls_uv,
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       (row << 3) + 4, col << 3,
-                                      &b->mv[2][1], 4, 4, w2, h2, 1);
+                                      &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
                         mc_chroma_dir(s, mc[4][b->filter][1],
                                       s->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
                                       s->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
                                       ref2->data[1], ref2->linesize[1],
                                       ref2->data[2], ref2->linesize[2], tref2,
                                       (row << 3) + 4, (col << 3) + 4,
-                                      &b->mv[3][1], 4, 4, w2, h2, 1);
+                                      &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
                     }
                 }
             }
@@ -404,7 +408,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
 
         mc_luma_dir(s, mc[bwl][b->filter][0], s->dst[0], ls_y,
                     ref1->data[0], ref1->linesize[0], tref1,
-                    row << 3, col << 3, &b->mv[0][0], bw, bh, w1, h1, 0);
+                    row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
         w1 = (w1 + s->ss_h) >> s->ss_h;
         h1 = (h1 + s->ss_v) >> s->ss_v;
         mc_chroma_dir(s, mc[bwl + s->ss_h][b->filter][0],
@@ -412,12 +416,12 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                       ref1->data[1], ref1->linesize[1],
                       ref1->data[2], ref1->linesize[2], tref1,
                       row << (3 - s->ss_v), col << (3 - s->ss_h),
-                      &b->mv[0][0], uvbw, uvbh, w1, h1, 0);
+                      &b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
 
         if (b->comp) {
             mc_luma_dir(s, mc[bwl][b->filter][1], s->dst[0], ls_y,
                         ref2->data[0], ref2->linesize[0], tref2,
-                        row << 3, col << 3, &b->mv[0][1], bw, bh, w2, h2, 1);
+                        row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
             w2 = (w2 + s->ss_h) >> s->ss_h;
             h2 = (h2 + s->ss_v) >> s->ss_v;
             mc_chroma_dir(s, mc[bwl + s->ss_h][b->filter][1],
@@ -425,7 +429,7 @@ static void FN(inter_pred)(AVCodecContext *ctx)
                           ref2->data[1], ref2->linesize[1],
                           ref2->data[2], ref2->linesize[2], tref2,
                           row << (3 - s->ss_v), col << (3 - s->ss_h),
-                          &b->mv[0][1], uvbw, uvbh, w2, h2, 1);
+                          &b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
         }
     }
 }