enabled alternative transform for inter blocks using ADST_ADST.

DST 32X32 included but not used.

Change-Id: I0183c48a7b43087562a9f048360c139414ec54e0

configure

@@ -256,7 +256,14 @@ EXPERIMENT_LIST="
     multiple_arf
     non420
     alpha
+    masked_interinter
+    interintra
+    masked_interintra
+    filterintra
+    ext_tx
+    ext_tx_dst32
 "
+
 CONFIG_LIST="
     external_build
     install_docs

test/partial_idct_test.cc

@@ -71,7 +71,7 @@ TEST_P(PartialIDctTest, ResultsMatch) {
     size = 32;
     break;
   default:
-    ASSERT_TRUE(0) << "Wrong Size!";
+    FAIL() << "Wrong Size!";
     break;
   }
   DECLARE_ALIGNED_ARRAY(16, int16_t, test_coef_block1, kMaxNumCoeffs);

test/test_libvpx.cc

@@ -22,8 +22,10 @@ extern void vp9_rtcd();
 }
 #include "third_party/googletest/src/include/gtest/gtest.h"
 
-static void append_gtest_filter(const char *str) {
+static void append_negative_gtest_filter(const char *str) {
   std::string filter = ::testing::FLAGS_gtest_filter;
+  // Negative patterns begin with one '-' followed by a ':' separated list.
+  if (filter.find('-') == std::string::npos) filter += '-';
   filter += str;
   ::testing::FLAGS_gtest_filter = filter;
 }
@@ -34,21 +36,21 @@ int main(int argc, char **argv) {
 #if ARCH_X86 || ARCH_X86_64
   const int simd_caps = x86_simd_caps();
   if (!(simd_caps & HAS_MMX))
-    append_gtest_filter(":-MMX/*");
+    append_negative_gtest_filter(":-MMX/*");
   if (!(simd_caps & HAS_SSE))
-    append_gtest_filter(":-SSE/*");
+    append_negative_gtest_filter(":-SSE/*");
   if (!(simd_caps & HAS_SSE2))
-    append_gtest_filter(":-SSE2/*");
+    append_negative_gtest_filter(":-SSE2/*");
   if (!(simd_caps & HAS_SSE3))
-    append_gtest_filter(":-SSE3/*");
+    append_negative_gtest_filter(":-SSE3/*");
   if (!(simd_caps & HAS_SSSE3))
-    append_gtest_filter(":-SSSE3/*");
+    append_negative_gtest_filter(":-SSSE3/*");
   if (!(simd_caps & HAS_SSE4_1))
-    append_gtest_filter(":-SSE4_1/*");
+    append_negative_gtest_filter(":-SSE4_1/*");
   if (!(simd_caps & HAS_AVX))
-    append_gtest_filter(":-AVX/*");
+    append_negative_gtest_filter(":AVX/*");
   if (!(simd_caps & HAS_AVX2))
-    append_gtest_filter(":-AVX2/*");
+    append_negative_gtest_filter(":AVX2/*");
 #endif
 
 #if !CONFIG_SHARED

vp8/encoder/onyx_if.c

@@ -1401,6 +1401,7 @@ static void update_layer_contexts (VP8_COMP *cpi)
         unsigned int i;
         double prev_layer_framerate=0;
 
+        assert(oxcf->number_of_layers <= VPX_TS_MAX_LAYERS);
         for (i=0; i<oxcf->number_of_layers; i++)
         {
             LAYER_CONTEXT *lc = &cpi->layer_context[i];
@@ -1623,6 +1624,12 @@ void vp8_change_config(VP8_COMP *cpi, VP8_CONFIG *oxcf)
         cpi->oxcf.maximum_buffer_size =
             rescale((int)cpi->oxcf.maximum_buffer_size,
                     cpi->oxcf.target_bandwidth, 1000);
+    // Under a configuration change, where maximum_buffer_size may change,
+    // keep buffer level clipped to the maximum allowed buffer size.
+    if (cpi->bits_off_target > cpi->oxcf.maximum_buffer_size) {
+      cpi->bits_off_target = cpi->oxcf.maximum_buffer_size;
+      cpi->buffer_level = cpi->bits_off_target;
+    }
 
     /* Set up frame rate and related parameters rate control values. */
     vp8_new_framerate(cpi, cpi->framerate);
@@ -2164,10 +2171,12 @@ void vp8_remove_compressor(VP8_COMP **ptr)
                                               8.0 / 1000.0  / time_encoded;
                         double samples = 3.0 / 2 * cpi->frames_in_layer[i] *
                                          lst_yv12->y_width * lst_yv12->y_height;
-                        double total_psnr = vp8_mse2psnr(samples, 255.0,
-                                                  cpi->total_error2[i]);
-                        double total_psnr2 = vp8_mse2psnr(samples, 255.0,
-                                                  cpi->total_error2_p[i]);
+                        double total_psnr =
+                            vp8_mse2psnr(samples, 255.0,
+                                         cpi->total_error2[i]);
+                        double total_psnr2 =
+                            vp8_mse2psnr(samples, 255.0,
+                                         cpi->total_error2_p[i]);
                         double total_ssim = 100 * pow(cpi->sum_ssim[i] /
                                                       cpi->sum_weights[i], 8.0);
 
@@ -2185,9 +2194,9 @@ void vp8_remove_compressor(VP8_COMP **ptr)
                     double samples = 3.0 / 2 * cpi->count *
                                         lst_yv12->y_width * lst_yv12->y_height;
                     double total_psnr = vp8_mse2psnr(samples, 255.0,
-                                                         cpi->total_sq_error);
+                                                     cpi->total_sq_error);
                     double total_psnr2 = vp8_mse2psnr(samples, 255.0,
-                                                         cpi->total_sq_error2);
+                                                      cpi->total_sq_error2);
                     double total_ssim = 100 * pow(cpi->summed_quality /
                                                       cpi->summed_weights, 8.0);
 
@@ -2516,8 +2525,8 @@ static void generate_psnr_packet(VP8_COMP *cpi)
     pkt.data.psnr.samples[3] = width * height;
 
     for (i = 0; i < 4; i++)
-        pkt.data.psnr.psnr[i] = vp8_mse2psnr(pkt.data.psnr.samples[i], 255.0,
-                                             (double)(pkt.data.psnr.sse[i]));
+      pkt.data.psnr.psnr[i] = vp8_mse2psnr(pkt.data.psnr.samples[i], 255.0,
+                                           (double)(pkt.data.psnr.sse[i]));
 
     vpx_codec_pkt_list_add(cpi->output_pkt_list, &pkt);
 }
@@ -2675,8 +2684,8 @@ static int resize_key_frame(VP8_COMP *cpi)
     VP8_COMMON *cm = &cpi->common;
 
     /* Do we need to apply resampling for one pass cbr.
-     * In one pass this is more limited than in two pass cbr
-     * The test and any change is only made one per key frame sequence
+     * In one pass this is more limited than in two pass cbr.
+     * The test and any change is only made once per key frame sequence.
      */
     if (cpi->oxcf.allow_spatial_resampling && (cpi->oxcf.end_usage == USAGE_STREAM_FROM_SERVER))
     {
@@ -2699,7 +2708,7 @@ static int resize_key_frame(VP8_COMP *cpi)
             cm->vert_scale = (cm->vert_scale > NORMAL) ? cm->vert_scale - 1 : NORMAL;
         }
 
-        /* Get the new hieght and width */
+        /* Get the new height and width */
         Scale2Ratio(cm->horiz_scale, &hr, &hs);
         Scale2Ratio(cm->vert_scale, &vr, &vs);
         new_width = ((hs - 1) + (cpi->oxcf.Width * hr)) / hs;
@@ -5063,6 +5072,7 @@ int vp8_get_compressed_data(VP8_COMP *cpi, unsigned int *frame_flags, unsigned l
                 unsigned int i;
 
                 /* Update frame rates for each layer */
+                assert(cpi->oxcf.number_of_layers <= VPX_TS_MAX_LAYERS);
                 for (i=0; i<cpi->oxcf.number_of_layers; i++)
                 {
                     LAYER_CONTEXT *lc = &cpi->layer_context[i];

vp9/common/vp9_blockd.h

@@ -120,9 +120,73 @@ static INLINE int mi_height_log2(BLOCK_SIZE sb_type) {
   return mi_height_log2_lookup[sb_type];
 }
 
+#if CONFIG_MASKED_INTERINTER
+#define MASK_BITS_SML   3
+#define MASK_BITS_MED   4
+#define MASK_BITS_BIG   5
+#define MASK_NONE      -1
+
+static inline int get_mask_bits(BLOCK_SIZE sb_type) {
+  if (sb_type < BLOCK_8X8)
+    return 0;
+  if (sb_type <= BLOCK_8X8)
+    return MASK_BITS_SML;
+  else if (sb_type <= BLOCK_32X32)
+    return MASK_BITS_MED;
+  else
+    return MASK_BITS_BIG;
+}
+#endif
+
+#if CONFIG_INTERINTRA
+static INLINE TX_SIZE intra_size_log2_for_interintra(int bs) {
+  switch (bs) {
+    case 4:
+      return TX_4X4;
+      break;
+    case 8:
+      return TX_8X8;
+      break;
+    case 16:
+      return TX_16X16;
+      break;
+    case 32:
+      return TX_32X32;
+      break;
+    default:
+      return TX_32X32;
+      break;
+  }
+}
+
+static INLINE int is_interintra_allowed(BLOCK_SIZE sb_type) {
+  return ((sb_type >= BLOCK_8X8) && (sb_type < BLOCK_64X64));
+}
+
+#if CONFIG_MASKED_INTERINTRA
+#define MASK_BITS_SML_INTERINTRA   3
+#define MASK_BITS_MED_INTERINTRA   4
+#define MASK_BITS_BIG_INTERINTRA   5
+#define MASK_NONE_INTERINTRA      -1
+static INLINE int get_mask_bits_interintra(BLOCK_SIZE sb_type) {
+  if (sb_type == BLOCK_4X4)
+     return 0;
+  if (sb_type <= BLOCK_8X8)
+    return MASK_BITS_SML_INTERINTRA;
+  else if (sb_type <= BLOCK_32X32)
+    return MASK_BITS_MED_INTERINTRA;
+  else
+    return MASK_BITS_BIG_INTERINTRA;
+}
+#endif
+#endif
+
 // This structure now relates to 8x8 block regions.
 typedef struct {
   MB_PREDICTION_MODE mode, uv_mode;
+#if CONFIG_FILTERINTRA
+  int filterbit, uv_filterbit;
+#endif
   MV_REFERENCE_FRAME ref_frame[2];
   TX_SIZE tx_size;
   int_mv mv[2];                // for each reference frame used
@@ -139,11 +203,30 @@ typedef struct {
 
   INTERPOLATION_TYPE interp_filter;
 
+#if CONFIG_EXT_TX
+  EXT_TX_TYPE ext_txfrm;
+#endif
+
   BLOCK_SIZE sb_type;
+#if CONFIG_MASKED_INTERINTER
+  int use_masked_interinter;
+  int mask_index;
+#endif
+#if CONFIG_INTERINTRA
+  MB_PREDICTION_MODE interintra_mode, interintra_uv_mode;
+#if CONFIG_MASKED_INTERINTRA
+  int interintra_mask_index;
+  int interintra_uv_mask_index;
+  int use_masked_interintra;
+#endif
+#endif
 } MB_MODE_INFO;
 
 typedef struct {
   MB_MODE_INFO mbmi;
+#if CONFIG_FILTERINTRA
+  int b_filter_info[4];
+#endif
   b_mode_info bmi[4];
 } MODE_INFO;
 
@@ -151,6 +234,17 @@ static INLINE int is_inter_block(const MB_MODE_INFO *mbmi) {
   return mbmi->ref_frame[0] > INTRA_FRAME;
 }
 
+#if CONFIG_FILTERINTRA
+static INLINE int
+is_filter_allowed(MB_PREDICTION_MODE mode) {
+  return 1;
+}
+
+static INLINE int is_filter_enabled(TX_SIZE txsize) {
+  return (txsize <= TX_32X32);
+}
+#endif
+
 static INLINE int has_second_ref(const MB_MODE_INFO *mbmi) {
   return mbmi->ref_frame[1] > INTRA_FRAME;
 }
@@ -267,6 +361,7 @@ static BLOCK_SIZE get_subsize(BLOCK_SIZE bsize, PARTITION_TYPE partition) {
 
 extern const TX_TYPE mode2txfm_map[MB_MODE_COUNT];
 
+#if !CONFIG_EXT_TX
 static INLINE TX_TYPE get_tx_type_4x4(PLANE_TYPE plane_type,
                                       const MACROBLOCKD *xd, int ib) {
   const MODE_INFO *const mi = xd->mi_8x8[0];
@@ -278,18 +373,85 @@ static INLINE TX_TYPE get_tx_type_4x4(PLANE_TYPE plane_type,
   return mode2txfm_map[mbmi->sb_type < BLOCK_8X8 ? mi->bmi[ib].as_mode
                                                  : mbmi->mode];
 }
+#else
+static INLINE TX_TYPE get_tx_type_4x4(PLANE_TYPE plane_type,
+                                      const MACROBLOCKD *xd, int ib) {
+  const MODE_INFO *const mi = xd->mi_8x8[0];
+  const MB_MODE_INFO *const mbmi = &mi->mbmi;
 
+  if (plane_type != PLANE_TYPE_Y || xd->lossless)
+    return DCT_DCT;
+
+  if (is_inter_block(mbmi)) {
+    if (mbmi->ext_txfrm == NORM)
+      return DCT_DCT;
+    else
+      return ADST_ADST;
+  }
+
+  return mode2txfm_map[mbmi->sb_type < BLOCK_8X8 ? mi->bmi[ib].as_mode
+                                                 : mbmi->mode];
+}
+#endif
+
+#if !CONFIG_EXT_TX
 static INLINE TX_TYPE get_tx_type_8x8(PLANE_TYPE plane_type,
                                       const MACROBLOCKD *xd) {
   return plane_type == PLANE_TYPE_Y ? mode2txfm_map[xd->mi_8x8[0]->mbmi.mode]
                                     : DCT_DCT;
 }
+#else
+static INLINE TX_TYPE get_tx_type_8x8(PLANE_TYPE plane_type,
+                                      const MACROBLOCKD *xd) {
+  const MODE_INFO *const mi = xd->mi_8x8[0];
+  const MB_MODE_INFO *const mbmi = &mi->mbmi;
+  if (!is_inter_block(mbmi) || plane_type != PLANE_TYPE_Y) {
+    return plane_type == PLANE_TYPE_Y ? mode2txfm_map[xd->mi_8x8[0]->mbmi.mode]
+                                    : DCT_DCT;
+  } else {
+    if (mbmi->ext_txfrm == NORM)
+      return DCT_DCT;
+    else
+      return ADST_ADST;
+  }
+}
+#endif
 
+#if !CONFIG_EXT_TX
 static INLINE TX_TYPE get_tx_type_16x16(PLANE_TYPE plane_type,
                                         const MACROBLOCKD *xd) {
   return plane_type == PLANE_TYPE_Y ? mode2txfm_map[xd->mi_8x8[0]->mbmi.mode]
                                     : DCT_DCT;
 }
+#else
+static INLINE TX_TYPE get_tx_type_16x16(PLANE_TYPE plane_type,
+                                        const MACROBLOCKD *xd) {
+  const MODE_INFO *const mi = xd->mi_8x8[0];
+  const MB_MODE_INFO *const mbmi = &mi->mbmi;
+  if (!is_inter_block(mbmi) || plane_type != PLANE_TYPE_Y) {
+    return plane_type == PLANE_TYPE_Y ? mode2txfm_map[xd->mi_8x8[0]->mbmi.mode]
+                                    : DCT_DCT;
+  } else {
+    if (mbmi->ext_txfrm == NORM)
+      return DCT_DCT;
+    else
+      return ADST_ADST;
+  }
+}
+static INLINE TX_TYPE get_tx_type_32x32(PLANE_TYPE plane_type,
+                                        const MACROBLOCKD *xd) {
+  const MODE_INFO *const mi = xd->mi_8x8[0];
+  const MB_MODE_INFO *const mbmi = &mi->mbmi;
+  if (!is_inter_block(mbmi) || plane_type != PLANE_TYPE_Y) {
+    return DCT_DCT;
+  } else {
+    if (mbmi->ext_txfrm == NORM)
+      return DCT_DCT;
+    else
+      return ADST_ADST;
+  }
+}
+#endif
 
 static void setup_block_dptrs(MACROBLOCKD *xd, int ss_x, int ss_y) {
   int i;

vp9/common/vp9_entropy.h

@@ -97,6 +97,10 @@ struct VP9Common;
 void vp9_default_coef_probs(struct VP9Common *cm);
 void vp9_adapt_coef_probs(struct VP9Common *cm);
 
+#if CONFIG_EXT_TX
+void vp9_init_ext_tx_prob(struct VP9Common *cm);
+#endif
+
 static INLINE void reset_skip_context(MACROBLOCKD *xd, BLOCK_SIZE bsize) {
   int i;
   for (i = 0; i < MAX_MB_PLANE; i++) {

vp9/common/vp9_entropymode.c

@@ -14,6 +14,26 @@
 #include "vp9/common/vp9_onyxc_int.h"
 #include "vp9/common/vp9_seg_common.h"
 
+#if CONFIG_MASKED_INTERINTER
+static const vp9_prob default_masked_interinter_prob[BLOCK_SIZES] = {
+    192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192
+};
+#endif
+
+#if CONFIG_INTERINTRA
+static const vp9_prob default_interintra_prob[BLOCK_SIZES] = {
+  192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192
+};
+#if CONFIG_MASKED_INTERINTRA
+static const vp9_prob default_masked_interintra_prob[BLOCK_SIZES] = {
+  192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192, 192
+};
+#endif
+#endif
+#if CONFIG_EXT_TX
+const vp9_prob default_ext_tx_prob = 204;  // 0.6 = 153, 0.7 = 178, 0.8 = 204
+#endif
+
 const vp9_prob vp9_kf_y_mode_prob[INTRA_MODES][INTRA_MODES][INTRA_MODES - 1] = {
   {  // above = dc
     { 137,  30,  42, 148, 151, 207,  70,  52,  91 },  // left = dc
@@ -161,6 +181,16 @@ static const vp9_prob default_if_uv_probs[INTRA_MODES][INTRA_MODES - 1] = {
   { 101,  21, 107, 181, 192, 103,  19,  67, 125 }   // y = tm
 };
 
+#if CONFIG_FILTERINTRA
+const vp9_prob vp9_default_filterintra_prob[TX_SIZES][INTRA_MODES] = {
+  // DC     V      H    D45   D135   D117   D153   D207    D63     TM
+  {153,   171,   147,   150,   129,   101,   100,   153,   132,   111},
+  {171,   173,   185,   131,    70,    53,    70,   148,   127,   114},
+  {175,   203,   213,    86,    45,    71,    41,   150,   125,   154},
+  {235,   230,   154,   202,   154,   205,    37,   128,     0,   202}
+};
+#endif
+
 const vp9_prob vp9_kf_partition_probs[PARTITION_CONTEXTS]
                                      [PARTITION_TYPES - 1] = {
   // 8x8 -> 4x4
@@ -327,6 +357,21 @@ void vp9_init_mbmode_probs(VP9_COMMON *cm) {
   cm->fc.tx_probs = default_tx_probs;
   vp9_copy(cm->fc.mbskip_probs, default_mbskip_probs);
   vp9_copy(cm->fc.inter_mode_probs, default_inter_mode_probs);
+#if CONFIG_FILTERINTRA
+  vp9_copy(cm->fc.filterintra_prob, vp9_default_filterintra_prob);
+#endif
+#if CONFIG_MASKED_INTERINTER
+  vp9_copy(cm->fc.masked_interinter_prob, default_masked_interinter_prob);
+#endif
+#if CONFIG_INTERINTRA
+  vp9_copy(cm->fc.interintra_prob, default_interintra_prob);
+#if CONFIG_MASKED_INTERINTRA
+  vp9_copy(cm->fc.masked_interintra_prob, default_masked_interintra_prob);
+#endif
+#endif
+#if CONFIG_EXT_TX
+  cm->fc.ext_tx_prob = default_ext_tx_prob;
+#endif
 }
 
 const vp9_tree_index vp9_switchable_interp_tree
@@ -386,9 +431,10 @@ void vp9_adapt_mode_probs(VP9_COMMON *cm) {
                 counts->partition[i], fc->partition_prob[i]);
 
   if (cm->mcomp_filter_type == SWITCHABLE) {
-    for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++)
+    for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++) {
       adapt_probs(vp9_switchable_interp_tree, pre_fc->switchable_interp_prob[i],
                   counts->switchable_interp[i], fc->switchable_interp_prob[i]);
+    }
   }
 
   if (cm->tx_mode == TX_MODE_SELECT) {
@@ -418,6 +464,48 @@ void vp9_adapt_mode_probs(VP9_COMMON *cm) {
   for (i = 0; i < MBSKIP_CONTEXTS; ++i)
     fc->mbskip_probs[i] = adapt_prob(pre_fc->mbskip_probs[i],
                                      counts->mbskip[i]);
+
+#if CONFIG_FILTERINTRA
+  for (i = 0; i < TX_SIZES; ++i)
+    for (j = 0; j < INTRA_MODES; ++j)
+      fc->filterintra_prob[i][j] = adapt_prob(pre_fc->filterintra_prob[i][j],
+                                   counts->filterintra[i][j]);
+#endif
+
+#if CONFIG_MASKED_INTERINTER
+  if (cm->use_masked_interinter) {
+    for (i = 0; i < BLOCK_SIZES; ++i) {
+      if (get_mask_bits(i))
+        fc->masked_interinter_prob[i] = adapt_prob
+                                      (pre_fc->masked_interinter_prob[i],
+                                       counts->masked_interinter[i]);
+    }
+  }
+#endif
+
+#if CONFIG_INTERINTRA
+  if (cm->use_interintra) {
+    for (i = 0; i < BLOCK_SIZES; ++i) {
+      if (is_interintra_allowed(i))
+        fc->interintra_prob[i] = adapt_prob(pre_fc->interintra_prob[i],
+                                            counts->interintra[i]);
+    }
+#if CONFIG_MASKED_INTERINTRA
+    if (cm->use_masked_interintra) {
+      for (i = 0; i < BLOCK_SIZES; ++i) {
+        if (is_interintra_allowed(i) && get_mask_bits_interintra(i))
+          fc->masked_interintra_prob[i] = adapt_prob(
+                                          pre_fc->masked_interintra_prob[i],
+                                          counts->masked_interintra[i]);
+      }
+    }
+#endif
+  }
+#endif
+
+#if CONFIG_EXT_TX
+  fc->ext_tx_prob = adapt_prob(pre_fc->ext_tx_prob, counts->ext_tx);
+#endif
 }
 
 static void set_default_lf_deltas(struct loopfilter *lf) {

vp9/common/vp9_entropymode.h

@@ -18,6 +18,10 @@
 #define SWITCHABLE_FILTERS 3   // number of switchable filters
 #define SWITCHABLE_FILTER_CONTEXTS (SWITCHABLE_FILTERS + 1)
 
+#if CONFIG_INTERINTRA
+#define SEPARATE_INTERINTRA_UV  0
+#endif
+
 // #define MODE_STATS
 
 struct VP9Common;

vp9/common/vp9_entropymv.c

@@ -234,3 +234,4 @@ void vp9_adapt_mv_probs(VP9_COMMON *cm, int allow_hp) {
 void vp9_init_mv_probs(VP9_COMMON *cm) {
   cm->fc.nmvc = default_nmv_context;
 }
+

vp9/common/vp9_enums.h

@@ -79,6 +79,14 @@ typedef enum {
   TX_TYPES = 4
 } TX_TYPE;
 
+#if CONFIG_EXT_TX
+typedef enum {
+  NORM   = 0,
+  ALT = 1,
+  EXT_TX_TYPES = 2
+} EXT_TX_TYPE;
+#endif
+
 typedef enum {
   UNKNOWN    = 0,
   BT_601     = 1,  // YUV

vp9/common/vp9_idct.c

@@ -1378,6 +1378,54 @@ void vp9_idct32x32_add(const int16_t *input, uint8_t *dest, int stride,
     vp9_idct32x32_1024_add(input, dest, stride);
 }
 
+#if CONFIG_EXT_TX
+static double tmp[32*32];
+static double tmp2[32*32];
+extern double dstmtx32[32*32];
+
+void vp9_idst_add(const int16_t *input, uint8_t *dest, int stride,
+                  int eob, int size) {
+  int i, j, k;
+
+  double *basis;
+  int factor = (size == 32) ? 4 : 8;
+
+  switch (size) {
+    case 32:
+      basis = dstmtx32;
+      break;
+    default:
+      assert(0);
+      break;
+  }
+
+  for (i = 0; i < size; i++) {
+    for (j = 0; j < size; j++) {
+      tmp[i*size+j] = 0;
+      for (k = 0; k < size; k++) {
+        tmp[i*size+j] += input[i*size+k] * basis[j*size+k];  // row
+      }
+    }
+  }
+
+  for (i = 0; i < size; i++) {
+    for (j = 0; j < size; j++) {
+      tmp2[i*size+j] = 0;
+      for (k = 0; k < size; k++) {
+        tmp2[i*size+j] += basis[i*size+k] * tmp[k*size+j];  // col
+      }
+      if (tmp2[i*size+j] >= 0)
+        dest[i*stride+j] = clip_pixel((int)dest[i*stride+j] +
+                                      (int)(tmp2[i*size+j] / factor + 0.5));
+      else
+        dest[i*stride+j] = clip_pixel((int)dest[i*stride+j] +
+                                      (int)(tmp2[i*size+j] / factor - 0.5));
+    }
+  }
+}
+
+#endif
+
 // iht
 void vp9_iht4x4_add(TX_TYPE tx_type, const int16_t *input, uint8_t *dest,
                     int stride, int eob) {

vp9/common/vp9_idct.h

@@ -95,6 +95,12 @@ void vp9_idct16x16_add(const int16_t *input, uint8_t *dest, int stride, int
 void vp9_idct32x32_add(const int16_t *input, uint8_t *dest, int stride,
                        int eob);
 
+#if CONFIG_EXT_TX
+void vp9_idst_add(const int16_t *input, uint8_t *dest, int stride,
+                  int eob, int size);
+#endif
+
+
 void vp9_iht4x4_add(TX_TYPE tx_type, const int16_t *input, uint8_t *dest,
                     int stride, int eob);
 void vp9_iht8x8_add(TX_TYPE tx_type, const int16_t *input, uint8_t *dest,

vp9/common/vp9_onyxc_int.h

@@ -58,6 +58,21 @@ typedef struct frame_contexts {
   struct tx_probs tx_probs;
   vp9_prob mbskip_probs[MBSKIP_CONTEXTS];
   nmv_context nmvc;
+#if CONFIG_FILTERINTRA
+  vp9_prob filterintra_prob[TX_SIZES][INTRA_MODES];
+#endif
+#if CONFIG_MASKED_INTERINTER
+  vp9_prob masked_interinter_prob[BLOCK_SIZES];
+#endif
+#if CONFIG_INTERINTRA
+  vp9_prob interintra_prob[BLOCK_SIZES];
+#if CONFIG_MASKED_INTERINTRA
+  vp9_prob masked_interintra_prob[BLOCK_SIZES];
+#endif
+#endif
+#if CONFIG_EXT_TX
+  vp9_prob ext_tx_prob;
+#endif
 } FRAME_CONTEXT;
 
 typedef struct {
@@ -77,6 +92,21 @@ typedef struct {
   struct tx_counts tx;
   unsigned int mbskip[MBSKIP_CONTEXTS][2];
   nmv_context_counts mv;
+#if CONFIG_FILTERINTRA
+  unsigned int filterintra[TX_SIZES][INTRA_MODES][2];
+#endif
+#if CONFIG_MASKED_INTERINTER
+  unsigned int masked_interinter[BLOCK_SIZES][2];
+#endif
+#if CONFIG_INTERINTRA
+  unsigned int interintra[BLOCK_SIZES][2];
+#if CONFIG_MASKED_INTERINTRA
+  unsigned int masked_interintra[BLOCK_SIZES][2];
+#endif
+#endif
+#if CONFIG_EXT_TX
+  unsigned int ext_tx[2];
+#endif
 } FRAME_COUNTS;
 
 
@@ -222,6 +252,15 @@ typedef struct VP9Common {
   int fb_lru;  // Flag telling if lru is on/off
   uint32_t *fb_idx_ref_lru;  // Frame buffer lru cache
   uint32_t fb_idx_ref_lru_count;
+#if CONFIG_MASKED_INTERINTER
+  int use_masked_interinter;
+#endif
+#if CONFIG_INTERINTRA
+  int use_interintra;
+#if CONFIG_MASKED_INTERINTRA
+  int use_masked_interintra;
+#endif
+#endif
 } VP9_COMMON;
 
 // ref == 0 => LAST_FRAME

vp9/common/vp9_reconinter.c

@@ -139,6 +139,302 @@ MV clamp_mv_to_umv_border_sb(const MACROBLOCKD *xd, const MV *src_mv,
   return clamped_mv;
 }
 
+#if CONFIG_MASKED_INTERINTER
+#define MASK_WEIGHT_BITS 6
+
+static int get_masked_weight(int m) {
+  #define SMOOTHER_LEN  32
+  static const uint8_t smoothfn[2 * SMOOTHER_LEN + 1] = {
+      0,  0,  0,  0,  0,  0,  0,  0,
+      0,  0,  0,  0,  0,  1,  1,  1,
+      1,  1,  2,  2,  3,  4,  5,  6,
+      8,  9, 12, 14, 17, 21, 24, 28,
+      32,
+      36, 40, 43, 47, 50, 52, 55, 56,
+      58, 59, 60, 61, 62, 62, 63, 63,
+      63, 63, 63, 64, 64, 64, 64, 64,
+      64, 64, 64, 64, 64, 64, 64, 64,
+  };
+  if (m < -SMOOTHER_LEN)
+    return 0;
+  else if (m > SMOOTHER_LEN)
+    return (1 << MASK_WEIGHT_BITS);
+  else
+    return smoothfn[m + SMOOTHER_LEN];
+}
+
+static int get_hard_mask(int m) {
+  return m > 0;
+}
+
+// Equation of line: f(x, y) = a[0]*(x - a[2]*w/4) + a[1]*(y - a[3]*h/4) = 0
+// The soft mask is obtained by computing f(x, y) and then calling
+// get_masked_weight(f(x, y)).
+static const int mask_params_sml[1 << MASK_BITS_SML][4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+};
+
+static const int mask_params_med_hgtw[1 << MASK_BITS_MED][4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+};
+
+static const int mask_params_med_hltw[1 << MASK_BITS_MED][4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+};
+
+static const int mask_params_med_heqw[1 << MASK_BITS_MED][4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  { 0,  2, 0, 1},
+  { 0, -2, 0, 1},
+  { 0,  2, 0, 3},
+  { 0, -2, 0, 3},
+  { 2,  0, 1, 0},
+  {-2,  0, 1, 0},
+  { 2,  0, 3, 0},
+  {-2,  0, 3, 0},
+};
+
+static const int mask_params_big_hgtw[1 << MASK_BITS_BIG][4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+
+  { 0,  2, 0, 1},
+  { 0, -2, 0, 1},
+  { 0,  2, 0, 2},
+  { 0, -2, 0, 2},
+  { 0,  2, 0, 3},
+  { 0, -2, 0, 3},
+  { 2,  0, 2, 0},
+  {-2,  0, 2, 0},
+};
+
+static const int mask_params_big_hltw[1 << MASK_BITS_BIG][4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+
+  { 0,  2, 0, 2},
+  { 0, -2, 0, 2},
+  { 2,  0, 1, 0},
+  {-2,  0, 1, 0},
+  { 2,  0, 2, 0},
+  {-2,  0, 2, 0},
+  { 2,  0, 3, 0},
+  {-2,  0, 3, 0},
+};
+
+static const int mask_params_big_heqw[1 << MASK_BITS_BIG][4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+
+  { 0,  2, 0, 1},
+  { 0, -2, 0, 1},
+  { 0,  2, 0, 3},
+  { 0, -2, 0, 3},
+  { 2,  0, 1, 0},
+  {-2,  0, 1, 0},
+  { 2,  0, 3, 0},
+  {-2,  0, 3, 0},
+};
+
+static const int *get_mask_params(int mask_index,
+                                  BLOCK_SIZE sb_type,
+                                  int h, int w) {
+  const int *a;
+  const int mask_bits = get_mask_bits(sb_type);
+
+  if (mask_index == MASK_NONE)
+    return NULL;
+
+  if (mask_bits == MASK_BITS_SML) {
+    a = mask_params_sml[mask_index];
+  } else if (mask_bits == MASK_BITS_MED) {
+    if (h > w)
+      a = mask_params_med_hgtw[mask_index];
+    else if (h < w)
+      a = mask_params_med_hltw[mask_index];
+    else
+      a = mask_params_med_heqw[mask_index];
+  } else if (mask_bits == MASK_BITS_BIG) {
+    if (h > w)
+      a = mask_params_big_hgtw[mask_index];
+    else if (h < w)
+      a = mask_params_big_hltw[mask_index];
+    else
+      a = mask_params_big_heqw[mask_index];
+  } else {
+    assert(0);
+  }
+  return a;
+}
+
+void vp9_generate_masked_weight(int mask_index,
+                                BLOCK_SIZE sb_type,
+                                int h, int w,
+                                uint8_t *mask, int stride) {
+  int i, j;
+  const int *a = get_mask_params(mask_index, sb_type, h, w);
+  if (!a) return;
+  for (i = 0; i < h; ++i)
+    for (j = 0; j < w; ++j) {
+      int x = (j - (a[2] * w) / 4);
+      int y = (i - (a[3] * h) / 4);
+      int m = a[0] * x + a[1] * y;
+      mask[i * stride + j] = get_masked_weight(m);
+    }
+}
+
+void vp9_generate_hard_mask(int mask_index, BLOCK_SIZE sb_type,
+                            int h, int w, uint8_t *mask, int stride) {
+  int i, j;
+  const int *a = get_mask_params(mask_index, sb_type, h, w);
+  if (!a) return;
+  for (i = 0; i < h; ++i)
+    for (j = 0; j < w; ++j) {
+      int x = (j - (a[2] * w) / 4);
+      int y = (i - (a[3] * h) / 4);
+      int m = a[0] * x + a[1] * y;
+      mask[i * stride + j] = get_hard_mask(m);
+    }
+}
+
+static void build_masked_compound(uint8_t *dst, int dst_stride,
+                                  uint8_t *dst2, int dst2_stride,
+                                  int mask_index, BLOCK_SIZE sb_type,
+                                  int h, int w) {
+  int i, j;
+  uint8_t mask[4096];
+  vp9_generate_masked_weight(mask_index, sb_type, h, w, mask, 64);
+  for (i = 0; i < h; ++i)
+    for (j = 0; j < w; ++j) {
+      int m = mask[i * 64 + j];
+      dst[i * dst_stride + j] =  (dst[i * dst_stride + j] * m +
+                                  dst2[i * dst2_stride + j] *
+                                  ((1 << MASK_WEIGHT_BITS) - m) +
+                                  (1 << (MASK_WEIGHT_BITS - 1))) >>
+                                 MASK_WEIGHT_BITS;
+    }
+}
+#endif
+
 // TODO(jkoleszar): In principle, pred_w, pred_h are unnecessary, as we could
 // calculate the subsampled BLOCK_SIZE, but that type isn't defined for
 // sizes smaller than 16x16 yet.
@@ -196,8 +492,21 @@ static void build_inter_predictors(MACROBLOCKD *xd, int plane, int block,
     pre += (scaled_mv.row >> SUBPEL_BITS) * pre_buf->stride
            + (scaled_mv.col >> SUBPEL_BITS);
 
+#if CONFIG_MASKED_INTERINTER
+    if (ref && get_mask_bits(mi->mbmi.sb_type)
+        && mi->mbmi.use_masked_interinter) {
+      uint8_t tmp_dst[4096];
+      inter_predictor(pre, pre_buf->stride, tmp_dst, 64,
+                     subpel_x, subpel_y, scale, w, h, 0, &xd->subpix, xs, ys);
+      build_masked_compound(dst, dst_buf->stride, tmp_dst, 64,
+                            mi->mbmi.mask_index, mi->mbmi.sb_type, h, w);
+    } else {
+#endif
     inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride,
                     subpel_x, subpel_y, scale, w, h, ref, &xd->subpix, xs, ys);
+#if CONFIG_MASKED_INTERINTER
+    }
+#endif
   }
 }
 
@@ -240,8 +549,22 @@ void vp9_build_inter_predictors_sbuv(MACROBLOCKD *xd, int mi_row, int mi_col,
 }
 void vp9_build_inter_predictors_sb(MACROBLOCKD *xd, int mi_row, int mi_col,
                                    BLOCK_SIZE bsize) {
+#if CONFIG_INTERINTRA
+  uint8_t *const y = xd->plane[0].dst.buf;
+  uint8_t *const u = xd->plane[1].dst.buf;
+  uint8_t *const v = xd->plane[2].dst.buf;
+  const int y_stride = xd->plane[0].dst.stride;
+  const int uv_stride = xd->plane[1].dst.stride;
+#endif
   build_inter_predictors_for_planes(xd, bsize, mi_row, mi_col, 0,
                                     MAX_MB_PLANE - 1);
+#if CONFIG_INTERINTRA
+  if (xd->mi_8x8[0]->mbmi.ref_frame[1] == INTRA_FRAME &&
+      is_interintra_allowed(xd->mi_8x8[0]->mbmi.sb_type)) {
+    vp9_build_interintra_predictors(xd, y, u, v,
+                                    y_stride, uv_stride, bsize);
+  }
+#endif
 }
 
 // TODO(jingning): This function serves as a placeholder for decoder prediction
@@ -360,8 +683,21 @@ static void dec_build_inter_predictors(MACROBLOCKD *xd, int plane, int block,
       }
     }
 
+#if CONFIG_MASKED_INTERINTER
+    if (ref && get_mask_bits(mi->mbmi.sb_type)
+        && mi->mbmi.use_masked_interinter) {
+      uint8_t tmp_dst[4096];
+      inter_predictor(buf_ptr, pre_buf->stride, tmp_dst, 64,
+                     subpel_x, subpel_y, scale, w, h, 0, &xd->subpix, xs, ys);
+      build_masked_compound(dst, dst_buf->stride, tmp_dst, 64,
+                            mi->mbmi.mask_index, mi->mbmi.sb_type, h, w);
+    } else {
+#endif
     inter_predictor(buf_ptr, pre_buf->stride, dst, dst_buf->stride, subpel_x,
                     subpel_y, scale, w, h, ref, &xd->subpix, xs, ys);
+#if CONFIG_MASKED_INTERINTER
+    }
+#endif
   }
 }
 
@@ -370,6 +706,13 @@ void vp9_dec_build_inter_predictors_sb(MACROBLOCKD *xd, int mi_row, int mi_col,
   int plane;
   const int mi_x = mi_col * MI_SIZE;
   const int mi_y = mi_row * MI_SIZE;
+#if CONFIG_INTERINTRA
+  uint8_t *const y = xd->plane[0].dst.buf;
+  uint8_t *const u = xd->plane[1].dst.buf;
+  uint8_t *const v = xd->plane[2].dst.buf;
+  const int y_stride = xd->plane[0].dst.stride;
+  const int uv_stride = xd->plane[1].dst.stride;
+#endif
   for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
     const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize,
                                                         &xd->plane[plane]);
@@ -390,6 +733,13 @@ void vp9_dec_build_inter_predictors_sb(MACROBLOCKD *xd, int mi_row, int mi_col,
                                  0, 0, bw, bh, mi_x, mi_y);
     }
   }
+#if CONFIG_INTERINTRA
+  if (xd->mi_8x8[0]->mbmi.ref_frame[1] == INTRA_FRAME &&
+      is_interintra_allowed(xd->mi_8x8[0]->mbmi.sb_type)) {
+    vp9_build_interintra_predictors(xd, y, u, v,
+                                    y_stride, uv_stride, bsize);
+  }
+#endif
 }
 
 // TODO(dkovalev: find better place for this function)

vp9/common/vp9_reconinter.h

@@ -99,4 +99,11 @@ static void set_scale_factors(MACROBLOCKD *xd, int ref0, int ref1,
 
 void vp9_setup_scale_factors(VP9_COMMON *cm, int i);
 
+#if CONFIG_MASKED_INTERINTER
+void vp9_generate_masked_weight(int mask_index, BLOCK_SIZE sb_type,
+                              int h, int w, uint8_t *mask, int stride);
+void vp9_generate_hard_mask(int mask_index, BLOCK_SIZE sb_type,
+                          int h, int w, uint8_t *mask, int stride);
+#endif
+
 #endif  // VP9_COMMON_VP9_RECONINTER_H_

vp9/common/vp9_reconintra.c

@@ -437,8 +437,226 @@ static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
   }
 }
 
+#if CONFIG_FILTERINTRA
+static void filter_intra_predictors_4tap(uint8_t *ypred_ptr, int y_stride,
+                                        int bs,
+                                        const uint8_t *yabove_row,
+                                        const uint8_t *yleft_col,
+                                        int mode) {
+  static const int prec_bits = 10;
+  static const int round_val = 511;
+
+  int k, r, c;
+  int pred[33][33];
+  int mean, ipred;
+
+  int taps4_4[10][4] = {
+      {735, 881, -537, -54},
+      {1005, 519, -488, -11},
+      {383, 990, -343, -6},
+      {442, 805, -542, 319},
+      {658, 616, -133, -116},
+      {875, 442, -141, -151},
+      {386, 741, -23, -80},
+      {390, 1027, -446, 51},
+      {679, 606, -523, 262},
+      {903, 922, -778, -23}
+  };
+  int taps4_8[10][4] = {
+      {648, 803, -444, 16},
+      {972, 620, -576, 7},
+      {561, 967, -499, -5},
+      {585, 762, -468, 144},
+      {596, 619, -182, -9},
+      {895, 459, -176, -153},
+      {557, 722, -126, -129},
+      {601, 839, -523, 105},
+      {562, 709, -499, 251},
+      {803, 872, -695, 43}
+  };
+  int taps4_16[10][4] = {
+      {423, 728, -347, 111},
+      {963, 685, -665, 23},
+      {281, 1024, -480, 216},
+      {640, 596, -437, 78},
+      {429, 669, -259, 99},
+      {740, 646, -415, 23},
+      {568, 771, -346, 40},
+      {404, 833, -486, 209},
+      {398, 712, -423, 307},
+      {939, 935, -887, 17}
+  };
+  int taps4_32[10][4] = {
+      {477, 737, -393, 150},
+      {881, 630, -546, 67},
+      {506, 984, -443, -20},
+      {114, 459, -270, 528},
+      {433, 528, 14, 3},
+      {837, 470, -301, -30},
+      {181, 777, 89, -107},
+      {-29, 716, -232, 259},
+      {589, 646, -495, 255},
+      {740, 884, -728, 77}
+  };
+
+  const int c1 = (bs >= 32) ? taps4_32[mode][0] : ((bs >= 16) ?
+      taps4_16[mode][0] : ((bs >= 8) ? taps4_8[mode][0] : taps4_4[mode][0]));
+  const int c2 = (bs >= 32) ? taps4_32[mode][1] : ((bs >= 16) ?
+      taps4_16[mode][1] : ((bs >= 8) ? taps4_8[mode][1] : taps4_4[mode][1]));
+  const int c3 = (bs >= 32) ? taps4_32[mode][2] : ((bs >= 16) ?
+      taps4_16[mode][2] : ((bs >= 8) ? taps4_8[mode][2] : taps4_4[mode][2]));
+  const int c4 = (bs >= 32) ? taps4_32[mode][3] : ((bs >= 16) ?
+      taps4_16[mode][3] : ((bs >= 8) ? taps4_8[mode][3] : taps4_4[mode][3]));
+
+  k = 0;
+  mean = 0;
+  while (k < bs) {
+    mean = mean + (int)yleft_col[k];
+    mean = mean + (int)yabove_row[k];
+    k++;
+  }
+  mean = (mean + bs) / (2 * bs);
+
+  for (r = 0; r < bs; r++)
+    pred[r + 1][0] = (int)yleft_col[r] - mean;
+
+  for (c = 0; c < 2 * bs + 1; c++)
+    pred[0][c] = (int)yabove_row[c - 1] - mean;
+
+  for (r = 1; r < bs + 1; r++)
+    for (c = 1; c < 2 * bs + 1 - r; c++) {
+      ipred = c1 * pred[r - 1][c] + c2 * pred[r][c - 1]
+                    + c3 * pred[r - 1][c - 1] + c4 * pred[r - 1][c + 1];
+      pred[r][c] = ipred < 0 ? -((-ipred + round_val) >> prec_bits) :
+                               ((ipred + round_val) >> prec_bits);
+    }
+
+  for (r = 0; r < bs; r++) {
+    for (c = 0; c < bs; c++) {
+      ipred = pred[r + 1][c + 1] + mean;
+      ypred_ptr[c] = clip_pixel(ipred);
+    }
+    ypred_ptr += y_stride;
+  }
+}
+
+static void build_filter_intra_predictors(const MACROBLOCKD *xd,
+                                          const uint8_t *ref, int ref_stride,
+                                          uint8_t *dst, int dst_stride,
+                                          MB_PREDICTION_MODE mode, TX_SIZE txsz,
+                                          int up_available, int left_available,
+                                          int right_available, int x, int y,
+                                          int plane) {
+  int i;
+  DECLARE_ALIGNED_ARRAY(16, uint8_t, left_col, 64);
+  DECLARE_ALIGNED_ARRAY(16, uint8_t, above_data, 128 + 16);
+  uint8_t *above_row = above_data + 16;
+  const uint8_t *const_above_row = above_row;
+  const int bs = 4 << txsz;
+  int frame_width, frame_height;
+  int x0, y0;
+  const struct macroblockd_plane *const pd = &xd->plane[plane];
+
+  // Get current frame pointer, width and height.
+  if (plane == 0) {
+    frame_width = xd->cur_buf->y_width;
+    frame_height = xd->cur_buf->y_height;
+  } else {
+    frame_width = xd->cur_buf->uv_width;
+    frame_height = xd->cur_buf->uv_height;
+  }
+
+  // Get block position in current frame.
+  x0 = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x)) + x;
+  y0 = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y)) + y;
+
+  // left
+  if (left_available) {
+    if (xd->mb_to_bottom_edge < 0) {
+      /* slower path if the block needs border extension */
+      if (y0 + bs <= frame_height) {
+        for (i = 0; i < bs; ++i)
+          left_col[i] = ref[i * ref_stride - 1];
+      } else {
+        const int extend_bottom = frame_height - y0;
+        for (i = 0; i < extend_bottom; ++i)
+          left_col[i] = ref[i * ref_stride - 1];
+        for (; i < bs; ++i)
+          left_col[i] = ref[(extend_bottom - 1) * ref_stride - 1];
+      }
+    } else {
+      /* faster path if the block does not need extension */
+      for (i = 0; i < bs; ++i)
+        left_col[i] = ref[i * ref_stride - 1];
+    }
+  } else {
+    vpx_memset(left_col, 129, bs);
+  }
+
+  // TODO(hkuang) do not extend 2*bs pixels for all modes.
+  // above
+  if (up_available) {
+    const uint8_t *above_ref = ref - ref_stride;
+    if (xd->mb_to_right_edge < 0) {
+      /* slower path if the block needs border extension */
+      if (x0 + 2 * bs <= frame_width) {
+        if (right_available && bs == 4) {
+          vpx_memcpy(above_row - 1, above_ref - 1, 2 * bs + 1);
+        } else {
+          vpx_memcpy(above_row - 1, above_ref - 1, bs + 1);
+          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+        }
+      } else if (x0 + bs <= frame_width) {
+        const int r = frame_width - x0;
+        if (right_available && bs == 4) {
+          vpx_memcpy(above_row - 1, above_ref - 1, r + 1);
+          vpx_memset(above_row + r, above_row[r - 1],
+                     x0 + 2 * bs - frame_width);
+        } else {
+          vpx_memcpy(above_row - 1, above_ref - 1, bs + 1);
+          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+        }
+      } else if (x0 <= frame_width) {
+        const int r = frame_width - x0;
+        if (right_available && bs == 4) {
+          vpx_memcpy(above_row - 1, above_ref - 1, r + 1);
+          vpx_memset(above_row + r, above_row[r - 1],
+                     x0 + 2 * bs - frame_width);
+        } else {
+          vpx_memcpy(above_row - 1, above_ref - 1, r + 1);
+          vpx_memset(above_row + r, above_row[r - 1],
+                     x0 + 2 * bs - frame_width);
+        }
+        above_row[-1] = left_available ? above_ref[-1] : 129;
+      }
+    } else {
+      /* faster path if the block does not need extension */
+      if (bs == 4 && right_available && left_available) {
+        const_above_row = above_ref;
+      } else {
+        vpx_memcpy(above_row, above_ref, bs);
+        if (bs == 4 && right_available)
+          vpx_memcpy(above_row + bs, above_ref + bs, bs);
+        else
+          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+        above_row[-1] = left_available ? above_ref[-1] : 129;
+      }
+    }
+  } else {
+    vpx_memset(above_row, 127, bs * 2);
+    above_row[-1] = 127;
+  }
+
+  filter_intra_predictors_4tap(dst, dst_stride, bs, const_above_row, left_col,
+                               mode);
+}
+#endif
+
 void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
                              TX_SIZE tx_size, int mode,
+#if CONFIG_FILTERINTRA
+                             int filterbit,
+#endif
                              const uint8_t *ref, int ref_stride,
                              uint8_t *dst, int dst_stride,
                              int aoff, int loff, int plane) {
@@ -449,8 +667,768 @@ void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
   const int have_right = ((block_idx & wmask) != wmask);
   const int x = aoff * 4;
   const int y = loff * 4;
+#if CONFIG_FILTERINTRA
+  const int filterflag = is_filter_allowed(mode) && is_filter_enabled(tx_size)
+                         && filterbit;
+#endif
 
   assert(bwl >= 0);
+#if CONFIG_FILTERINTRA
+  if (!filterflag) {
+#endif
   build_intra_predictors(xd, ref, ref_stride, dst, dst_stride, mode, tx_size,
                          have_top, have_left, have_right, x, y, plane);
+#if CONFIG_FILTERINTRA
+  } else {
+    build_filter_intra_predictors(xd, ref, ref_stride, dst, dst_stride, mode,
+                        tx_size, have_top, have_left, have_right, x, y, plane);
+  }
+#endif
 }
+
+#if CONFIG_INTERINTRA
+#if CONFIG_MASKED_INTERINTRA
+#define MASK_WEIGHT_BITS_INTERINTRA 6
+
+static int get_masked_weight_interintra(int m) {
+  #define SMOOTHER_LEN_INTERINTRA  32
+  static const uint8_t smoothfn[2 * SMOOTHER_LEN_INTERINTRA + 1] = {
+      0,  0,  0,  0,  0,  0,  0,  0,
+      0,  0,  0,  0,  0,  1,  1,  1,
+      1,  1,  2,  2,  3,  4,  5,  6,
+      8,  9, 12, 14, 17, 21, 24, 28,
+      32,
+      36, 40, 43, 47, 50, 52, 55, 56,
+      58, 59, 60, 61, 62, 62, 63, 63,
+      63, 63, 63, 64, 64, 64, 64, 64,
+      64, 64, 64, 64, 64, 64, 64, 64,
+  };
+  if (m < -SMOOTHER_LEN_INTERINTRA)
+    return 0;
+  else if (m > SMOOTHER_LEN_INTERINTRA)
+    return (1 << MASK_WEIGHT_BITS_INTERINTRA);
+  else
+    return smoothfn[m + SMOOTHER_LEN_INTERINTRA];
+}
+
+static int get_hard_mask_interintra(int m) {
+  return m > 0;
+}
+
+// Equation of line: f(x, y) = a[0]*(x - a[2]*w/4) + a[1]*(y - a[3]*h/4) = 0
+// The soft mask is obtained by computing f(x, y) and then calling
+// get_masked_weight(f(x, y)).
+static const int mask_params_sml_interintra[1 << MASK_BITS_SML_INTERINTRA]
+                                            [4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+};
+
+static const int mask_params_med_hgtw_interintra[1 << MASK_BITS_MED_INTERINTRA]
+                                                 [4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+};
+
+static const int mask_params_med_hltw_interintra[1 << MASK_BITS_MED_INTERINTRA]
+                                                 [4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+};
+
+static const int mask_params_med_heqw_interintra[1 << MASK_BITS_MED_INTERINTRA]
+                                                 [4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  { 0,  2, 0, 1},
+  { 0, -2, 0, 1},
+  { 0,  2, 0, 3},
+  { 0, -2, 0, 3},
+  { 2,  0, 1, 0},
+  {-2,  0, 1, 0},
+  { 2,  0, 3, 0},
+  {-2,  0, 3, 0},
+};
+
+static const int mask_params_big_hgtw_interintra[1 << MASK_BITS_BIG_INTERINTRA]
+                                                 [4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+
+  { 0,  2, 0, 1},
+  { 0, -2, 0, 1},
+  { 0,  2, 0, 2},
+  { 0, -2, 0, 2},
+  { 0,  2, 0, 3},
+  { 0, -2, 0, 3},
+  { 2,  0, 2, 0},
+  {-2,  0, 2, 0},
+};
+
+static const int mask_params_big_hltw_interintra[1 << MASK_BITS_BIG_INTERINTRA]
+                                                 [4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+
+  { 0,  2, 0, 2},
+  { 0, -2, 0, 2},
+  { 2,  0, 1, 0},
+  {-2,  0, 1, 0},
+  { 2,  0, 2, 0},
+  {-2,  0, 2, 0},
+  { 2,  0, 3, 0},
+  {-2,  0, 3, 0},
+};
+
+static const int mask_params_big_heqw_interintra[1 << MASK_BITS_BIG_INTERINTRA]
+                                                 [4] = {
+  {-1,  2, 2, 2},
+  { 1, -2, 2, 2},
+  {-2,  1, 2, 2},
+  { 2, -1, 2, 2},
+  { 2,  1, 2, 2},
+  {-2, -1, 2, 2},
+  { 1,  2, 2, 2},
+  {-1, -2, 2, 2},
+
+  {-1,  2, 2, 1},
+  { 1, -2, 2, 1},
+  {-1,  2, 2, 3},
+  { 1, -2, 2, 3},
+  { 1,  2, 2, 1},
+  {-1, -2, 2, 1},
+  { 1,  2, 2, 3},
+  {-1, -2, 2, 3},
+
+  {-2,  1, 1, 2},
+  { 2, -1, 1, 2},
+  {-2,  1, 3, 2},
+  { 2, -1, 3, 2},
+  { 2,  1, 1, 2},
+  {-2, -1, 1, 2},
+  { 2,  1, 3, 2},
+  {-2, -1, 3, 2},
+
+  { 0,  2, 0, 1},
+  { 0, -2, 0, 1},
+  { 0,  2, 0, 3},
+  { 0, -2, 0, 3},
+  { 2,  0, 1, 0},
+  {-2,  0, 1, 0},
+  { 2,  0, 3, 0},
+  {-2,  0, 3, 0},
+};
+
+static const int *get_mask_params_interintra(int mask_index,
+                                             BLOCK_SIZE sb_type,
+                                             int h, int w) {
+  const int *a;
+  const int mask_bits = get_mask_bits_interintra(sb_type);
+
+  if (mask_index == MASK_NONE_INTERINTRA)
+    return NULL;
+
+  if (mask_bits == MASK_BITS_SML_INTERINTRA) {
+    a = mask_params_sml_interintra[mask_index];
+  } else if (mask_bits == MASK_BITS_MED_INTERINTRA) {
+    if (h > w)
+      a = mask_params_med_hgtw_interintra[mask_index];
+    else if (h < w)
+      a = mask_params_med_hltw_interintra[mask_index];
+    else
+      a = mask_params_med_heqw_interintra[mask_index];
+  } else if (mask_bits == MASK_BITS_BIG_INTERINTRA) {
+    if (h > w)
+      a = mask_params_big_hgtw_interintra[mask_index];
+    else if (h < w)
+      a = mask_params_big_hltw_interintra[mask_index];
+    else
+      a = mask_params_big_heqw_interintra[mask_index];
+  } else {
+    assert(0);
+  }
+  return a;
+}
+
+void vp9_generate_masked_weight_interintra(int mask_index,
+                                           BLOCK_SIZE sb_type,
+                                           int h, int w,
+                                           uint8_t *mask, int stride) {
+  int i, j;
+  const int *a = get_mask_params_interintra(mask_index, sb_type, h, w);
+  if (!a) return;
+  for (i = 0; i < h; ++i)
+    for (j = 0; j < w; ++j) {
+      int x = (j - (a[2] * w) / 4);
+      int y = (i - (a[3] * h) / 4);
+      int m = a[0] * x + a[1] * y;
+      mask[i * stride + j] = get_masked_weight_interintra(m);
+    }
+}
+
+void vp9_generate_hard_mask_interintra(int mask_index, BLOCK_SIZE sb_type,
+                            int h, int w, uint8_t *mask, int stride) {
+  int i, j;
+  const int *a = get_mask_params_interintra(mask_index, sb_type, h, w);
+  if (!a) return;
+  for (i = 0; i < h; ++i)
+    for (j = 0; j < w; ++j) {
+      int x = (j - (a[2] * w) / 4);
+      int y = (i - (a[3] * h) / 4);
+      int m = a[0] * x + a[1] * y;
+      mask[i * stride + j] = get_hard_mask_interintra(m);
+    }
+}
+#endif
+
+static void combine_interintra(MB_PREDICTION_MODE mode,
+#if CONFIG_MASKED_INTERINTRA
+                               int use_masked_interintra,
+                               int mask_index,
+                               BLOCK_SIZE bsize,
+#endif
+                               uint8_t *interpred,
+                               int interstride,
+                               uint8_t *intrapred,
+                               int intrastride,
+                               int bw, int bh) {
+  static const int scale_bits = 8;
+  static const int scale_max = 256;
+  static const int scale_round = 127;
+  static const int weights1d[64] = {
+      128, 125, 122, 119, 116, 114, 111, 109,
+      107, 105, 103, 101,  99,  97,  96,  94,
+       93,  91,  90,  89,  88,  86,  85,  84,
+       83,  82,  81,  81,  80,  79,  78,  78,
+       77,  76,  76,  75,  75,  74,  74,  73,
+       73,  72,  72,  71,  71,  71,  70,  70,
+       70,  70,  69,  69,  69,  69,  68,  68,
+       68,  68,  68,  67,  67,  67,  67,  67,
+  };
+
+  int size = MAX(bw, bh);
+  int size_scale = (size >= 64 ? 1 :
+                    size == 32 ? 2 :
+                    size == 16 ? 4 :
+                    size == 8  ? 8 : 16);
+  int i, j;
+
+#if CONFIG_MASKED_INTERINTRA
+  uint8_t mask[4096];
+  if (use_masked_interintra && get_mask_bits_interintra(bsize))
+    vp9_generate_masked_weight_interintra(mask_index, bsize, bh, bw, mask, bw);
+#endif
+
+  switch (mode) {
+    case V_PRED:
+      for (i = 0; i < bh; ++i) {
+        for (j = 0; j < bw; ++j) {
+          int k = i * interstride + j;
+          int scale = weights1d[i * size_scale];
+#if CONFIG_MASKED_INTERINTRA
+          int m = mask[i * bw + j];
+          if (use_masked_interintra && get_mask_bits_interintra(bsize))
+              interpred[k] = (intrapred[i * intrastride + j] * m +
+                              interpred[k] *
+                              ((1 << MASK_WEIGHT_BITS_INTERINTRA) - m) +
+                              (1 << (MASK_WEIGHT_BITS_INTERINTRA - 1))) >>
+                              MASK_WEIGHT_BITS_INTERINTRA;
+          else
+#endif
+          interpred[k] =
+              ((scale_max - scale) * interpred[k] +
+                  scale * intrapred[i * intrastride + j] + scale_round)
+                  >> scale_bits;
+        }
+      }
+     break;
+
+    case H_PRED:
+      for (i = 0; i < bh; ++i) {
+        for (j = 0; j < bw; ++j) {
+          int k = i * interstride + j;
+          int scale = weights1d[j * size_scale];
+#if CONFIG_MASKED_INTERINTRA
+          int m = mask[i * bw + j];
+          if (use_masked_interintra && get_mask_bits_interintra(bsize))
+              interpred[k] = (intrapred[i * intrastride + j] * m +
+                              interpred[k] *
+                              ((1 << MASK_WEIGHT_BITS_INTERINTRA) - m) +
+                              (1 << (MASK_WEIGHT_BITS_INTERINTRA - 1))) >>
+                              MASK_WEIGHT_BITS_INTERINTRA;
+          else
+#endif
+          interpred[k] =
+              ((scale_max - scale) * interpred[k] +
+                  scale * intrapred[i * intrastride + j] + scale_round)
+                  >> scale_bits;
+        }
+      }
+     break;
+
+    case D63_PRED:
+    case D117_PRED:
+      for (i = 0; i < bh; ++i) {
+        for (j = 0; j < bw; ++j) {
+          int k = i * interstride + j;
+          int scale = (weights1d[i * size_scale] * 3 +
+                       weights1d[j * size_scale]) >> 2;
+#if CONFIG_MASKED_INTERINTRA
+          int m = mask[i * bw + j];
+          if (use_masked_interintra && get_mask_bits_interintra(bsize))
+              interpred[k] = (intrapred[i * intrastride + j] * m +
+                              interpred[k] *
+                              ((1 << MASK_WEIGHT_BITS_INTERINTRA) - m) +
+                              (1 << (MASK_WEIGHT_BITS_INTERINTRA - 1))) >>
+                              MASK_WEIGHT_BITS_INTERINTRA;
+          else
+#endif
+          interpred[k] =
+              ((scale_max - scale) * interpred[k] +
+                  scale * intrapred[i * intrastride + j] + scale_round)
+                  >> scale_bits;
+        }
+      }
+     break;
+
+    case D207_PRED:
+    case D153_PRED:
+      for (i = 0; i < bh; ++i) {
+        for (j = 0; j < bw; ++j) {
+          int k = i * interstride + j;
+          int scale = (weights1d[j * size_scale] * 3 +
+                       weights1d[i * size_scale]) >> 2;
+#if CONFIG_MASKED_INTERINTRA
+          int m = mask[i * bw + j];
+          if (use_masked_interintra && get_mask_bits_interintra(bsize))
+              interpred[k] = (intrapred[i * intrastride + j] * m +
+                              interpred[k] *
+                              ((1 << MASK_WEIGHT_BITS_INTERINTRA) - m) +
+                              (1 << (MASK_WEIGHT_BITS_INTERINTRA - 1))) >>
+                              MASK_WEIGHT_BITS_INTERINTRA;
+          else
+#endif
+          interpred[k] =
+              ((scale_max - scale) * interpred[k] +
+                  scale * intrapred[i * intrastride + j] + scale_round)
+                  >> scale_bits;
+        }
+      }
+     break;
+
+    case D135_PRED:
+      for (i = 0; i < bh; ++i) {
+        for (j = 0; j < bw; ++j) {
+          int k = i * interstride + j;
+          int scale = weights1d[(i < j ? i : j) * size_scale];
+#if CONFIG_MASKED_INTERINTRA
+          int m = mask[i * bw + j];
+          if (use_masked_interintra && get_mask_bits_interintra(bsize))
+              interpred[k] = (intrapred[i * intrastride + j] * m +
+                              interpred[k] *
+                              ((1 << MASK_WEIGHT_BITS_INTERINTRA) - m) +
+                              (1 << (MASK_WEIGHT_BITS_INTERINTRA - 1))) >>
+                              MASK_WEIGHT_BITS_INTERINTRA;
+          else
+#endif
+          interpred[k] =
+              ((scale_max - scale) * interpred[k] +
+                  scale * intrapred[i * intrastride + j] + scale_round)
+                  >> scale_bits;
+        }
+      }
+     break;
+
+    case D45_PRED:
+      for (i = 0; i < bh; ++i) {
+        for (j = 0; j < bw; ++j) {
+          int k = i * interstride + j;
+          int scale = (weights1d[i * size_scale] +
+                       weights1d[j * size_scale]) >> 1;
+#if CONFIG_MASKED_INTERINTRA
+          int m = mask[i * bw + j];
+          if (use_masked_interintra && get_mask_bits_interintra(bsize))
+              interpred[k] = (intrapred[i * intrastride + j] * m +
+                              interpred[k] *
+                              ((1 << MASK_WEIGHT_BITS_INTERINTRA) - m) +
+                              (1 << (MASK_WEIGHT_BITS_INTERINTRA - 1))) >>
+                              MASK_WEIGHT_BITS_INTERINTRA;
+          else
+#endif
+          interpred[k] =
+              ((scale_max - scale) * interpred[k] +
+                  scale * intrapred[i * intrastride + j] + scale_round)
+                  >> scale_bits;
+        }
+      }
+     break;
+
+    case TM_PRED:
+    case DC_PRED:
+    default:
+      for (i = 0; i < bh; ++i) {
+        for (j = 0; j < bw; ++j) {
+          int k = i * interstride + j;
+#if CONFIG_MASKED_INTERINTRA
+          int m = mask[i * bw + j];
+          if (use_masked_interintra && get_mask_bits_interintra(bsize))
+              interpred[k] = (intrapred[i * intrastride + j] * m +
+                              interpred[k] *
+                              ((1 << MASK_WEIGHT_BITS_INTERINTRA) - m) +
+                              (1 << (MASK_WEIGHT_BITS_INTERINTRA - 1))) >>
+                              MASK_WEIGHT_BITS_INTERINTRA;
+          else
+#endif
+          interpred[k] = (interpred[k] + intrapred[i * intrastride + j]) >> 1;
+        }
+      }
+      break;
+  }
+}
+
+// Intra predictor for the second square block in interintra prediction.
+// Prediction of the first block (in pred_ptr) will be used to generate half of
+// the boundary values.
+static void build_intra_predictors_for_2nd_block_interintra
+                                  (const MACROBLOCKD *xd, const uint8_t *ref,
+                                   int ref_stride, uint8_t *dst, int dst_stride,
+                                   MB_PREDICTION_MODE mode, TX_SIZE tx_size,
+                                   int up_available, int left_available,
+                                   int right_available, int bwltbh,
+                                   int x, int y, int plane) {
+  int i;
+  DECLARE_ALIGNED_ARRAY(16, uint8_t, left_col, 64);
+  DECLARE_ALIGNED_ARRAY(16, uint8_t, above_data, 128 + 16);
+  uint8_t *above_row = above_data + 16;
+  const uint8_t *const_above_row = above_row;
+  const int bs = 4 << tx_size;
+  int frame_width, frame_height;
+  int x0, y0;
+  const struct macroblockd_plane *const pd = &xd->plane[plane];
+  const uint8_t *ref_fi;
+  int ref_stride_fi;
+
+  // 127 127 127 .. 127 127 127 127 127 127
+  // 129  A   B  ..  Y   Z
+  // 129  C   D  ..  W   X
+  // 129  E   F  ..  U   V
+  // 129  G   H  ..  S   T   T   T   T   T
+  // ..
+
+  once(init_intra_pred_fn_ptrs);
+
+  // Get current frame pointer, width and height.
+  if (plane == 0) {
+    frame_width = xd->cur_buf->y_width;
+    frame_height = xd->cur_buf->y_height;
+  } else {
+    frame_width = xd->cur_buf->uv_width;
+    frame_height = xd->cur_buf->uv_height;
+  }
+
+  // Get block position in current frame.
+  x0 = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x)) + x;
+  y0 = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y)) + y;
+
+  // left
+  if (left_available) {
+    if (bwltbh) {
+      ref_fi = ref;
+      ref_stride_fi = ref_stride;
+    } else {
+      ref_fi = dst;
+      ref_stride_fi = dst_stride;
+    }
+
+    if (xd->mb_to_bottom_edge < 0) {
+      /* slower path if the block needs border extension */
+      if (y0 + bs <= frame_height) {
+        for (i = 0; i < bs; ++i)
+          left_col[i] = ref_fi[i * ref_stride_fi - 1];
+      } else {
+        const int extend_bottom = frame_height - y0;
+        for (i = 0; i < extend_bottom; ++i)
+          left_col[i] = ref_fi[i * ref_stride_fi - 1];
+        for (; i < bs; ++i)
+          left_col[i] = ref_fi[(extend_bottom - 1) * ref_stride_fi - 1];
+      }
+    } else {
+      /* faster path if the block does not need extension */
+      for (i = 0; i < bs; ++i)
+        left_col[i] = ref_fi[i * ref_stride_fi - 1];
+    }
+  } else {
+    vpx_memset(left_col, 129, bs);
+  }
+
+  // TODO(hkuang) do not extend 2*bs pixels for all modes.
+  // above
+  if (up_available) {
+    const uint8_t *above_ref;
+    if (bwltbh) {
+      ref_fi = dst;
+      ref_stride_fi = dst_stride;
+      above_row[-1] = left_available ? ref[-ref_stride-1] : 129;
+    } else {
+      ref_fi = ref;
+      ref_stride_fi = ref_stride;
+      above_row[-1] = ref[-ref_stride-1];
+    }
+    above_ref = ref_fi - ref_stride_fi;
+    if (xd->mb_to_right_edge < 0) {
+      /* slower path if the block needs border extension */
+      if (x0 + 2 * bs <= frame_width) {
+        if (right_available && bs == 4) {
+          vpx_memcpy(above_row, above_ref, 2 * bs);
+        } else {
+          vpx_memcpy(above_row, above_ref, bs);
+          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+        }
+      } else if (x0 + bs <= frame_width) {
+        const int r = frame_width - x0;
+        if (right_available && bs == 4) {
+          vpx_memcpy(above_row, above_ref, r);
+          vpx_memset(above_row + r, above_row[r - 1],
+                     x0 + 2 * bs - frame_width);
+        } else {
+          vpx_memcpy(above_row, above_ref, bs);
+          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+        }
+      } else if (x0 <= frame_width) {
+        const int r = frame_width - x0;
+        if (right_available && bs == 4) {
+          vpx_memcpy(above_row, above_ref, r);
+          vpx_memset(above_row + r, above_row[r - 1],
+                     x0 + 2 * bs - frame_width);
+        } else {
+          vpx_memcpy(above_row, above_ref, r);
+          vpx_memset(above_row + r, above_row[r - 1],
+                     x0 + 2 * bs - frame_width);
+        }
+        // above_row[-1] = left_available ? above_ref[-1] : 129;
+      }
+    } else {
+      /* faster path if the block does not need extension */
+      if (bs == 4 && right_available && left_available) {
+        const_above_row = above_ref;
+      } else {
+        vpx_memcpy(above_row, above_ref, bs);
+        if (bs == 4 && right_available)
+          vpx_memcpy(above_row + bs, above_ref + bs, bs);
+        else
+          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+        // above_row[-1] = left_available ? above_ref[-1] : 129;
+      }
+    }
+  } else {
+    vpx_memset(above_row, 127, bs * 2);
+    above_row[-1] = 127;
+  }
+
+  // predict
+  if (mode == DC_PRED) {
+    dc_pred[left_available][up_available][tx_size](dst, dst_stride,
+                                                   const_above_row, left_col);
+  } else {
+    pred[mode][tx_size](dst, dst_stride, const_above_row, left_col);
+  }
+}
+
+// Break down rectangular intra prediction for joint spatio-temporal prediction
+// into two square intra predictions.
+static void build_intra_predictors_for_interintra(MACROBLOCKD *xd,
+                                           uint8_t *src, int src_stride,
+                                           uint8_t *pred_ptr, int stride,
+                                           MB_PREDICTION_MODE mode,
+                                           int bw, int bh,
+                                           int up_available, int left_available,
+                                           int right_available, int plane) {
+  if (bw == bh) {
+    build_intra_predictors(xd, src, src_stride, pred_ptr, stride,
+                           mode, intra_size_log2_for_interintra(bw),
+                           up_available, left_available, right_available,
+                           0, 0, plane);
+  } else if (bw < bh) {
+    uint8_t *src_bottom = src + bw * src_stride;
+    uint8_t *pred_ptr_bottom = pred_ptr + bw * stride;
+    build_intra_predictors(xd, src, src_stride, pred_ptr, stride,
+                           mode, intra_size_log2_for_interintra(bw),
+                           up_available, left_available, right_available,
+                           0, 0, plane);
+    build_intra_predictors_for_2nd_block_interintra(xd, src_bottom, src_stride,
+                           pred_ptr_bottom, stride,
+                           mode, intra_size_log2_for_interintra(bw),
+                           up_available, left_available, 0, 1,
+                           0, bw, plane);
+  } else {
+    uint8_t *src_right = src + bh;
+    uint8_t *pred_ptr_right = pred_ptr + bh;
+    build_intra_predictors(xd, src, src_stride, pred_ptr, stride,
+                           mode, intra_size_log2_for_interintra(bh),
+                           up_available, left_available, 1,
+                           0, 0, plane);
+    build_intra_predictors_for_2nd_block_interintra(xd, src_right, src_stride,
+                           pred_ptr_right, stride,
+                           mode, intra_size_log2_for_interintra(bh),
+                           up_available, left_available, right_available, 0,
+                           bh, 0, plane);
+  }
+}
+
+void vp9_build_interintra_predictors_sby(MACROBLOCKD *xd,
+                                         uint8_t *ypred,
+                                         int ystride,
+                                         BLOCK_SIZE bsize) {
+  int bw = 4 << b_width_log2(bsize);
+  int bh = 4 << b_height_log2(bsize);
+  uint8_t intrapredictor[4096];
+  build_intra_predictors_for_interintra(
+      xd, xd->plane[0].dst.buf, xd->plane[0].dst.stride,
+      intrapredictor, bw,
+      xd->mi_8x8[0]->mbmi.interintra_mode, bw, bh,
+      xd->up_available, xd->left_available, 0, 0);
+  combine_interintra(xd->mi_8x8[0]->mbmi.interintra_mode,
+#if CONFIG_MASKED_INTERINTRA
+                     xd->mi_8x8[0]->mbmi.use_masked_interintra,
+                     xd->mi_8x8[0]->mbmi.interintra_mask_index,
+                     bsize,
+#endif
+                     ypred, ystride, intrapredictor, bw, bw, bh);
+}
+
+void vp9_build_interintra_predictors_sbuv(MACROBLOCKD *xd,
+                                          uint8_t *upred,
+                                          uint8_t *vpred,
+                                          int uvstride,
+                                          BLOCK_SIZE bsize) {
+  int bwl = b_width_log2(bsize), bw = 2 << bwl;
+  int bhl = b_height_log2(bsize), bh = 2 << bhl;
+  uint8_t uintrapredictor[1024];
+  uint8_t vintrapredictor[1024];
+  build_intra_predictors_for_interintra(
+      xd, xd->plane[1].dst.buf, xd->plane[1].dst.stride,
+      uintrapredictor, bw,
+      xd->mi_8x8[0]->mbmi.interintra_uv_mode, bw, bh,
+      xd->up_available, xd->left_available, 0, 1);
+  build_intra_predictors_for_interintra(
+      xd, xd->plane[2].dst.buf, xd->plane[1].dst.stride,
+      vintrapredictor, bw,
+      xd->mi_8x8[0]->mbmi.interintra_uv_mode, bw, bh,
+      xd->up_available, xd->left_available, 0, 2);
+  combine_interintra(xd->mi_8x8[0]->mbmi.interintra_uv_mode,
+#if CONFIG_MASKED_INTERINTRA
+                     xd->mi_8x8[0]->mbmi.use_masked_interintra,
+                     xd->mi_8x8[0]->mbmi.interintra_uv_mask_index,
+                     bsize,
+#endif
+                     upred, uvstride, uintrapredictor, bw, bw, bh);
+  combine_interintra(xd->mi_8x8[0]->mbmi.interintra_uv_mode,
+#if CONFIG_MASKED_INTERINTRA
+                     xd->mi_8x8[0]->mbmi.use_masked_interintra,
+                     xd->mi_8x8[0]->mbmi.interintra_uv_mask_index,
+                     bsize,
+#endif
+                     vpred, uvstride, vintrapredictor, bw, bw, bh);
+}
+
+void vp9_build_interintra_predictors(MACROBLOCKD *xd,
+                                     uint8_t *ypred,
+                                     uint8_t *upred,
+                                     uint8_t *vpred,
+                                     int ystride, int uvstride,
+                                     BLOCK_SIZE bsize) {
+  vp9_build_interintra_predictors_sby(xd, ypred, ystride, bsize);
+  vp9_build_interintra_predictors_sbuv(xd, upred, vpred, uvstride, bsize);
+}
+#endif

vp9/common/vp9_reconintra.h

@@ -16,7 +16,26 @@
 
 void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
                              TX_SIZE tx_size, int mode,
+#if CONFIG_FILTERINTRA
+                             int filterbit,
+#endif
                              const uint8_t *ref, int ref_stride,
                              uint8_t *dst, int dst_stride,
                              int aoff, int loff, int plane);
 #endif  // VP9_COMMON_VP9_RECONINTRA_H_
+
+#if CONFIG_INTERINTRA
+void vp9_build_interintra_predictors(MACROBLOCKD *xd,
+                                     uint8_t *ypred,
+                                     uint8_t *upred,
+                                     uint8_t *vpred,
+                                     int ystride,
+                                     int uvstride,
+                                     BLOCK_SIZE bsize);
+#if CONFIG_MASKED_INTERINTRA
+void vp9_generate_masked_weight_interintra(int mask_index,
+                                           BLOCK_SIZE sb_type,
+                                           int h, int w,
+                                           uint8_t *mask, int stride);
+#endif
+#endif

vp9/common/vp9_rtcd_defs.sh

@@ -543,6 +543,113 @@ specialize vp9_sad4x8_avg $sse_x86inc
 prototype unsigned int vp9_sad4x4_avg "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *second_pred, unsigned int max_sad"
 specialize vp9_sad4x4_avg $sse_x86inc
 
+if [ "$CONFIG_MASKED_INTERINTER" = "yes" ] || ([ "$CONFIG_MASKED_INTERINTRA" = "yes" ] && [ "$CONFIG_INTERINTRA" = "yes" ]); then
+prototype int vp9_masked_diamond_search_sad "struct macroblock *x, uint8_t *mask, int mask_stride, struct mv *ref_mv, struct mv *best_mv, int search_param, int sad_per_bit, int *num00, struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv, int is_second"
+specialize vp9_masked_diamond_search_sad
+
+prototype int vp9_masked_refining_search_sad "struct macroblock *x, uint8_t *mask, int mask_stride, struct mv *ref_mv, int sad_per_bit, int distance, struct vp9_variance_vtable *fn_ptr, DEC_MVCOSTS, const struct mv *center_mv, int is_second"
+specialize vp9_masked_refining_search_sad
+
+prototype unsigned int vp9_masked_variance32x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance32x16
+
+prototype unsigned int vp9_masked_variance16x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance16x32
+
+prototype unsigned int vp9_masked_variance64x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance64x32
+
+prototype unsigned int vp9_masked_variance32x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance32x64
+
+prototype unsigned int vp9_masked_variance32x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance32x32
+
+prototype unsigned int vp9_masked_variance64x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance64x64
+
+prototype unsigned int vp9_masked_variance16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance16x16
+
+prototype unsigned int vp9_masked_variance16x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance16x8
+
+prototype unsigned int vp9_masked_variance8x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance8x16
+
+prototype unsigned int vp9_masked_variance8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance8x8
+
+prototype unsigned int vp9_masked_variance4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_variance4x4
+
+prototype unsigned int vp9_masked_sub_pixel_variance64x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance64x64
+
+prototype unsigned int vp9_masked_sub_pixel_variance32x64 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance32x64
+
+prototype unsigned int vp9_masked_sub_pixel_variance64x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance64x32
+
+prototype unsigned int vp9_masked_sub_pixel_variance32x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance32x16
+
+prototype unsigned int vp9_masked_sub_pixel_variance16x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance16x32
+
+prototype unsigned int vp9_masked_sub_pixel_variance32x32 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance32x32
+
+prototype unsigned int vp9_masked_sub_pixel_variance16x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance16x16
+
+prototype unsigned int vp9_masked_sub_pixel_variance8x16 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance8x16
+
+prototype unsigned int vp9_masked_sub_pixel_variance16x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance16x8
+
+prototype unsigned int vp9_masked_sub_pixel_variance8x8 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance8x8
+
+prototype unsigned int vp9_masked_sub_pixel_variance4x4 "const uint8_t *src_ptr, int source_stride, int xoffset, int  yoffset, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int *sse"
+specialize vp9_masked_sub_pixel_variance4x4
+
+prototype unsigned int vp9_masked_sad64x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad64x64
+
+prototype unsigned int vp9_masked_sad32x64 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad32x64
+
+prototype unsigned int vp9_masked_sad64x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad64x32
+
+prototype unsigned int vp9_masked_sad32x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad32x16
+
+prototype unsigned int vp9_masked_sad16x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad16x32
+
+prototype unsigned int vp9_masked_sad32x32 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad32x32
+
+prototype unsigned int vp9_masked_sad16x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad16x16
+
+prototype unsigned int vp9_masked_sad16x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad16x8
+
+prototype unsigned int vp9_masked_sad8x16 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad8x16
+
+prototype unsigned int vp9_masked_sad8x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad8x8
+
+prototype unsigned int vp9_masked_sad4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int  ref_stride, const uint8_t *msk_ptr, int msk_stride, unsigned int max_sad"
+specialize vp9_masked_sad4x4
+fi
+
 prototype unsigned int vp9_variance_halfpixvar16x16_h "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
 specialize vp9_variance_halfpixvar16x16_h $sse2_x86inc
 

vp9/decoder/vp9_decodeframe.c

@@ -140,15 +140,18 @@ static void read_reference_mode_probs(VP9_COMMON *cm, vp9_reader *r) {
     for (i = 0; i < COMP_INTER_CONTEXTS; i++)
       vp9_diff_update_prob(r, &cm->fc.comp_inter_prob[i]);
 
-  if (cm->reference_mode != COMPOUND_REFERENCE)
+  if (cm->reference_mode != COMPOUND_REFERENCE) {
     for (i = 0; i < REF_CONTEXTS; i++) {
       vp9_diff_update_prob(r, &cm->fc.single_ref_prob[i][0]);
       vp9_diff_update_prob(r, &cm->fc.single_ref_prob[i][1]);
     }
+  }
 
-  if (cm->reference_mode != SINGLE_REFERENCE)
-    for (i = 0; i < REF_CONTEXTS; i++)
+  if (cm->reference_mode != SINGLE_REFERENCE) {
+    for (i = 0; i < REF_CONTEXTS; i++) {
       vp9_diff_update_prob(r, &cm->fc.comp_ref_prob[i]);
+    }
+  }
 }
 
 static void update_mv_probs(vp9_prob *p, int n, vp9_reader *r) {
@@ -261,8 +264,19 @@ static void inverse_transform_block(MACROBLOCKD* xd, int plane, int block,
         vp9_iht16x16_add(tx_type, dqcoeff, dst, stride, eob);
         break;
       case TX_32X32:
+#if !CONFIG_EXT_TX
         tx_type = DCT_DCT;
         vp9_idct32x32_add(dqcoeff, dst, stride, eob);
+#else
+        tx_type = get_tx_type_32x32(plane_type, xd);
+        if (tx_type == DCT_DCT)
+          vp9_idct32x32_add(dqcoeff, dst, stride, eob);
+        else if (tx_type == ADST_ADST)
+          // yongzhe: temporarily signaled as ADST_ADST
+          vp9_idst_add(dqcoeff, dst, stride, eob, 32);
+        else
+          assert(0 && "Invalid transform type");
+#endif
         break;
       default:
         assert(0 && "Invalid transform size");
@@ -301,11 +315,24 @@ static void predict_and_reconstruct_intra_block(int plane, int block,
           : mi->mbmi.uv_mode;
   int x, y;
   uint8_t *dst;
+#if CONFIG_FILTERINTRA
+  int fbit;
+  if (plane == 0)
+    if (mi->mbmi.sb_type < BLOCK_8X8)
+      fbit = mi->b_filter_info[block];
+    else
+      fbit = is_filter_enabled(tx_size) ? mi->mbmi.filterbit : 0;
+  else
+    fbit = is_filter_enabled(tx_size) ? mi->mbmi.uv_filterbit : 0;
+#endif
   txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &x, &y);
   dst = &pd->dst.buf[4 * y * pd->dst.stride + 4 * x];
 
   vp9_predict_intra_block(xd, block >> (tx_size << 1),
                           b_width_log2(plane_bsize), tx_size, mode,
+#if CONFIG_FILTERINTRA
+                          fbit,
+#endif
                           dst, pd->dst.stride, dst, pd->dst.stride,
                           x, y, plane);
 
@@ -1261,6 +1288,10 @@ static int read_compressed_header(VP9D_COMP *pbi, const uint8_t *data,
     nmv_context *const nmvc = &fc->nmvc;
     int i, j;
 
+#if CONFIG_EXT_TX
+    vp9_diff_update_prob(&r, &fc->ext_tx_prob);
+#endif
+
     read_inter_mode_probs(fc, &r);
 
     if (cm->mcomp_filter_type == SWITCHABLE)
@@ -1272,6 +1303,50 @@ static int read_compressed_header(VP9D_COMP *pbi, const uint8_t *data,
     cm->reference_mode = read_reference_mode(cm, &r);
     read_reference_mode_probs(cm, &r);
 
+#if CONFIG_MASKED_INTERINTER
+    if (cm->reference_mode != SINGLE_REFERENCE) {
+      cm->use_masked_interinter = vp9_read_bit(&r);
+      if (cm->use_masked_interinter) {
+        for (i = 0; i < BLOCK_SIZES; i++) {
+          if (get_mask_bits(i)) {
+            vp9_diff_update_prob(&r, &fc->masked_interinter_prob[i]);
+          }
+        }
+      }
+    } else {
+      cm->use_masked_interinter = 0;
+    }
+#endif
+
+#if CONFIG_INTERINTRA
+    if (cm->reference_mode != COMPOUND_REFERENCE) {
+      cm->use_interintra = vp9_read_bit(&r);
+      if (cm->use_interintra) {
+        for (i = 0; i < BLOCK_SIZES; i++) {
+          if (is_interintra_allowed(i)) {
+            vp9_diff_update_prob(&r, &fc->interintra_prob[i]);
+          }
+        }
+#if CONFIG_MASKED_INTERINTRA
+        cm->use_masked_interintra = vp9_read_bit(&r);
+        if (cm->use_masked_interintra) {
+          for (i = 0; i < BLOCK_SIZES; i++) {
+            if (is_interintra_allowed(i) && get_mask_bits_interintra(i))
+              vp9_diff_update_prob(&r, &fc->masked_interintra_prob[i]);
+          }
+        }
+      } else {
+        cm->use_masked_interintra = 0;
+#endif
+      }
+    } else {
+      cm->use_interintra = 0;
+#if CONFIG_MASKED_INTERINTRA
+      cm->use_masked_interintra = 0;
+#endif
+    }
+#endif
+
     for (j = 0; j < BLOCK_SIZE_GROUPS; j++)
       for (i = 0; i < INTRA_MODES - 1; ++i)
         vp9_diff_update_prob(&r, &fc->y_mode_prob[j][i]);
@@ -1319,6 +1394,10 @@ static void debug_check_frame_counts(const VP9_COMMON *const cm) {
                  sizeof(cm->counts.eob_branch)));
   assert(!memcmp(cm->counts.switchable_interp, zero_counts.switchable_interp,
                  sizeof(cm->counts.switchable_interp)));
+#if CONFIG_EXT_TX
+  assert(!memcmp(cm->counts.ext_tx, zero_counts.ext_tx,
+                 sizeof(cm->counts.ext_tx)));
+#endif
   assert(!memcmp(cm->counts.inter_mode, zero_counts.inter_mode,
                  sizeof(cm->counts.inter_mode)));
   assert(!memcmp(cm->counts.intra_inter, zero_counts.intra_inter,

vp9/decoder/vp9_decodemv.c

@@ -179,6 +179,13 @@ static void read_intra_frame_mode_info(VP9_COMMON *const cm,
     const MB_PREDICTION_MODE A = above_block_mode(mi, above_mi, 0);
     const MB_PREDICTION_MODE L = left_block_mode(mi, left_mi, 0);
     mbmi->mode = read_intra_mode(r, vp9_kf_y_mode_prob[A][L]);
+#if CONFIG_FILTERINTRA
+    if (is_filter_enabled(mbmi->tx_size) && is_filter_allowed(mbmi->mode))
+      mbmi->filterbit = vp9_read(r,
+          cm->fc.filterintra_prob[mbmi->tx_size][mbmi->mode]);
+    else
+      mbmi->filterbit = 0;
+#endif
   } else {
     // Only 4x4, 4x8, 8x4 blocks
     const int num_4x4_w = num_4x4_blocks_wide_lookup[bsize];  // 1 or 2
@@ -197,13 +204,35 @@ static void read_intra_frame_mode_info(VP9_COMMON *const cm,
           mi->bmi[ib + 2].as_mode = b_mode;
         if (num_4x4_w == 2)
           mi->bmi[ib + 1].as_mode = b_mode;
+#if CONFIG_FILTERINTRA
+        if (is_filter_allowed(b_mode))
+          mi->b_filter_info[ib] = vp9_read(r,
+                                 cm->fc.filterintra_prob[0][b_mode]);
+        else
+          mi->b_filter_info[ib] = 0;
+        if (num_4x4_h == 2)
+          mi->b_filter_info[ib + 2] = mi->b_filter_info[ib];
+        if (num_4x4_w == 2)
+          mi->b_filter_info[ib + 1] = mi->b_filter_info[ib];
+#endif
       }
     }
 
     mbmi->mode = mi->bmi[3].as_mode;
+#if CONFIG_FILTERINTRA
+    mbmi->filterbit = mi->b_filter_info[3];
+#endif
   }
 
   mbmi->uv_mode = read_intra_mode(r, vp9_kf_uv_mode_prob[mbmi->mode]);
+#if CONFIG_FILTERINTRA
+  if (is_filter_enabled(get_uv_tx_size(mbmi)) &&
+      is_filter_allowed(mbmi->uv_mode))
+    mbmi->uv_filterbit = vp9_read(r,
+        cm->fc.filterintra_prob[get_uv_tx_size(mbmi)][mbmi->uv_mode]);
+  else
+    mbmi->uv_filterbit = 0;
+#endif
 }
 
 static int read_mv_component(vp9_reader *r,
@@ -334,6 +363,15 @@ static void read_intra_block_mode_info(VP9_COMMON *const cm, MODE_INFO *mi,
 
   if (bsize >= BLOCK_8X8) {
     mbmi->mode = read_intra_mode_y(cm, r, size_group_lookup[bsize]);
+#if CONFIG_FILTERINTRA
+    if (is_filter_allowed(mbmi->mode) && is_filter_enabled(mbmi->tx_size)) {
+      mbmi->filterbit = vp9_read(r,
+          cm->fc.filterintra_prob[mbmi->tx_size][mbmi->mode]);
+      cm->counts.filterintra[mbmi->tx_size][mbmi->mode][mbmi->filterbit]++;
+    } else {
+      mbmi->filterbit = 0;
+    }
+#endif
   } else {
      // Only 4x4, 4x8, 8x4 blocks
      const int num_4x4_w = num_4x4_blocks_wide_lookup[bsize];  // 1 or 2
@@ -349,12 +387,40 @@ static void read_intra_block_mode_info(VP9_COMMON *const cm, MODE_INFO *mi,
            mi->bmi[ib + 2].as_mode = b_mode;
          if (num_4x4_w == 2)
            mi->bmi[ib + 1].as_mode = b_mode;
+#if CONFIG_FILTERINTRA
+         if (is_filter_allowed(b_mode)) {
+           mi->b_filter_info[ib] = vp9_read(r,
+               cm->fc.filterintra_prob[0][b_mode]);
+           cm->counts.filterintra[0][b_mode][mi->b_filter_info[ib]]++;
+         } else {
+           mi->b_filter_info[ib] = 0;
+         }
+
+         if (num_4x4_h == 2)
+           mi->b_filter_info[ib + 2] = mi->b_filter_info[ib];
+         if (num_4x4_w == 2)
+           mi->b_filter_info[ib + 1] = mi->b_filter_info[ib];
+#endif
       }
     }
     mbmi->mode = mi->bmi[3].as_mode;
+#if CONFIG_FILTERINTRA
+    mbmi->filterbit = mi->b_filter_info[3];
+#endif
   }
 
   mbmi->uv_mode = read_intra_mode_uv(cm, r, mbmi->mode);
+#if CONFIG_FILTERINTRA
+  if (is_filter_allowed(mbmi->uv_mode) &&
+      is_filter_enabled(get_uv_tx_size(mbmi))) {
+    mbmi->uv_filterbit = vp9_read(r,
+        cm->fc.filterintra_prob[get_uv_tx_size(mbmi)][mbmi->uv_mode]);
+    cm->counts.filterintra[get_uv_tx_size(mbmi)]
+                           [mbmi->uv_mode][mbmi->uv_filterbit]++;
+  } else {
+    mbmi->uv_filterbit = 0;
+  }
+#endif
 }
 
 static INLINE int assign_mv(VP9_COMMON *cm, MB_PREDICTION_MODE mode,
@@ -428,6 +494,23 @@ static void read_inter_block_mode_info(VP9_COMMON *const cm,
   int inter_mode_ctx, ref, is_compound;
 
   read_ref_frames(cm, xd, r, mbmi->segment_id, mbmi->ref_frame);
+
+#if CONFIG_EXT_TX
+  assert(is_inter_block(mbmi));
+#if CONFIG_EXT_TX_DST32
+  if (
+#else
+  if (mbmi->tx_size <= TX_16X16 &&
+#endif
+      !mbmi->skip_coeff) {
+    mbmi->ext_txfrm = vp9_read(r, cm->fc.ext_tx_prob);
+    if (!cm->frame_parallel_decoding_mode)
+      ++cm->counts.ext_tx[mbmi->ext_txfrm];
+  } else {
+    mbmi->ext_txfrm = NORM;
+  }
+#endif
+
   is_compound = has_second_ref(mbmi);
 
   for (ref = 0; ref < 1 + is_compound; ++ref) {
@@ -462,6 +545,44 @@ static void read_inter_block_mode_info(VP9_COMMON *const cm,
                       ? read_switchable_filter_type(cm, xd, r)
                       : cm->mcomp_filter_type;
 
+#if CONFIG_INTERINTRA
+    if ((cm->use_interintra) &&
+        is_interintra_allowed(bsize) &&
+        is_inter_mode(mbmi->mode) &&
+        (mbmi->ref_frame[1] <= INTRA_FRAME)) {
+      mbmi->ref_frame[1] = vp9_read(r, cm->fc.interintra_prob[bsize]) ?
+                           INTRA_FRAME : NONE;
+      cm->counts.interintra[bsize][mbmi->ref_frame[1] == INTRA_FRAME]++;
+#if CONFIG_MASKED_INTERINTRA
+      mbmi->use_masked_interintra = 0;
+#endif
+      if (mbmi->ref_frame[1] == NONE)
+        mbmi->interintra_mode = -1;
+      if (mbmi->ref_frame[1] == INTRA_FRAME) {
+        mbmi->interintra_mode =
+            read_intra_mode_y(cm, r, size_group_lookup[bsize]);
+#if SEPARATE_INTERINTRA_UV
+        mbmi->interintra_uv_mode =
+            read_intra_mode_uv(cm, r, mbmi->interintra_mode);
+#else
+        mbmi->interintra_uv_mode = mbmi->interintra_mode;
+#endif
+#if CONFIG_MASKED_INTERINTRA
+        if (cm->use_masked_interintra && get_mask_bits_interintra(bsize)) {
+          mbmi->use_masked_interintra = vp9_read(r,
+                                          cm->fc.masked_interintra_prob[bsize]);
+          cm->counts.masked_interintra[bsize][mbmi->use_masked_interintra]++;
+          if (mbmi->use_masked_interintra) {
+            mbmi->interintra_mask_index = vp9_read_literal(r,
+                                               get_mask_bits_interintra(bsize));
+            mbmi->interintra_uv_mask_index = mbmi->interintra_mask_index;
+          }
+        }
+#endif
+      }
+    }
+#endif
+
   if (bsize < BLOCK_8X8) {
     const int num_4x4_w = num_4x4_blocks_wide_lookup[bsize];  // 1 or 2
     const int num_4x4_h = num_4x4_blocks_high_lookup[bsize];  // 1 or 2
@@ -504,6 +625,21 @@ static void read_inter_block_mode_info(VP9_COMMON *const cm,
                                 best, nearest, nearmv,
                                 is_compound, allow_hp, r);
   }
+#if CONFIG_MASKED_INTERINTER
+  mbmi->use_masked_interinter = 0;
+  if (cm->use_masked_interinter &&
+      cm->reference_mode != SINGLE_REFERENCE &&
+      is_inter_mode(mbmi->mode) &&
+      get_mask_bits(bsize) &&
+      mbmi->ref_frame[1] > INTRA_FRAME) {
+    mbmi->use_masked_interinter =
+        vp9_read(r, cm->fc.masked_interinter_prob[bsize]);
+    cm->counts.masked_interinter[bsize][mbmi->use_masked_interinter]++;
+    if (mbmi->use_masked_interinter) {
+      mbmi->mask_index = vp9_read_literal(r, get_mask_bits(bsize));
+    }
+  }
+#endif
 }
 
 static void read_inter_frame_mode_info(VP9_COMMON *const cm,

vp9/encoder/vp9_bitstream.c

@@ -307,6 +307,12 @@ static void pack_inter_mode_mvs(VP9_COMP *cpi, MODE_INFO *m, vp9_writer *bc) {
 
     if (bsize >= BLOCK_8X8) {
       write_intra_mode(bc, mode, cm->fc.y_mode_prob[size_group_lookup[bsize]]);
+#if CONFIG_FILTERINTRA
+      if (is_filter_allowed(mode) && is_filter_enabled(mi->tx_size)) {
+        vp9_write(bc, mi->filterbit,
+                  cm->fc.filterintra_prob[mi->tx_size][mode]);
+      }
+#endif
     } else {
       int idx, idy;
       const int num_4x4_blocks_wide = num_4x4_blocks_wide_lookup[bsize];
@@ -315,15 +321,42 @@ static void pack_inter_mode_mvs(VP9_COMP *cpi, MODE_INFO *m, vp9_writer *bc) {
         for (idx = 0; idx < 2; idx += num_4x4_blocks_wide) {
           const MB_PREDICTION_MODE bm = m->bmi[idy * 2 + idx].as_mode;
           write_intra_mode(bc, bm, cm->fc.y_mode_prob[0]);
+#if CONFIG_FILTERINTRA
+          if (is_filter_allowed(bm)) {
+            vp9_write(bc, m->b_filter_info[idy * 2 + idx],
+                      cm->fc.filterintra_prob[0][bm]);
+          }
+#endif
         }
       }
     }
     write_intra_mode(bc, mi->uv_mode, cm->fc.uv_mode_prob[mode]);
+#if CONFIG_FILTERINTRA
+    if (is_filter_allowed(mi->uv_mode) &&
+        is_filter_enabled(get_uv_tx_size(mi))) {
+      vp9_write(bc, mi->uv_filterbit,
+                cm->fc.filterintra_prob[get_uv_tx_size(mi)][mi->uv_mode]);
+    }
+#endif
   } else {
     vp9_prob *mv_ref_p;
     encode_ref_frame(cpi, bc);
     mv_ref_p = cpi->common.fc.inter_mode_probs[mi->mode_context[rf]];
 
+#if CONFIG_EXT_TX
+    assert(is_inter_block(mi));
+#if CONFIG_EXT_TX_DST32
+    if (
+#else
+    if (mi->tx_size <= TX_16X16 &&
+#endif
+        !mi->skip_coeff &&
+        !vp9_segfeature_active(&cm->seg, mi->segment_id, SEG_LVL_SKIP)) {
+//      printf("enc: writing bit\n");
+      vp9_write(bc, mi->ext_txfrm, cm->fc.ext_tx_prob);
+    }
+#endif
+
 #ifdef ENTROPY_STATS
     active_section = 3;
 #endif
@@ -345,6 +378,38 @@ static void pack_inter_mode_mvs(VP9_COMP *cpi, MODE_INFO *m, vp9_writer *bc) {
       assert(mi->interp_filter == cm->mcomp_filter_type);
     }
 
+#if CONFIG_INTERINTRA
+    if ((cm->use_interintra) &&
+        cpi->common.reference_mode != COMPOUND_REFERENCE &&
+        is_interintra_allowed(bsize) &&
+        is_inter_mode(mode) &&
+        (mi->ref_frame[1] <= INTRA_FRAME)) {
+        vp9_write(bc, mi->ref_frame[1] == INTRA_FRAME,
+                  cm->fc.interintra_prob[bsize]);
+        if (mi->ref_frame[1] == NONE)
+          mi->interintra_mode = -1;
+        if (mi->ref_frame[1] == INTRA_FRAME) {
+          write_intra_mode(bc, mi->interintra_mode,
+                           cm->fc.y_mode_prob[size_group_lookup[bsize]]);
+#if SEPARATE_INTERINTRA_UV
+          write_intra_mode(bc, mi->interintra_uv_mode,
+                           cm->fc.uv_mode_prob[mi->interintra_mode]);
+#endif
+#if CONFIG_MASKED_INTERINTRA
+        if (get_mask_bits_interintra(bsize) &&
+            cm->use_masked_interintra) {
+          vp9_write(bc, mi->use_masked_interintra,
+                    cm->fc.masked_interintra_prob[bsize]);
+          if (mi->use_masked_interintra) {
+            vp9_write_literal(bc, mi->interintra_mask_index,
+                              get_mask_bits_interintra(bsize));
+          }
+        }
+#endif
+      }
+    }
+#endif
+
     if (bsize < BLOCK_8X8) {
       const int num_4x4_blocks_wide = num_4x4_blocks_wide_lookup[bsize];
       const int num_4x4_blocks_high = num_4x4_blocks_high_lookup[bsize];
@@ -381,6 +446,19 @@ static void pack_inter_mode_mvs(VP9_COMP *cpi, MODE_INFO *m, vp9_writer *bc) {
         vp9_encode_mv(cpi, bc, &mi->mv[1].as_mv,
                       &mi->best_mv[1].as_mv, nmvc, allow_hp);
     }
+#if CONFIG_MASKED_INTERINTER
+  if (cpi->common.use_masked_interinter &&
+      cpi->common.reference_mode != SINGLE_REFERENCE &&
+      is_inter_mode(mode) &&
+      get_mask_bits(mi->sb_type) &&
+      mi->ref_frame[1] > INTRA_FRAME) {
+    vp9_write(bc, mi->use_masked_interinter,
+              cpi->common.fc.masked_interinter_prob[bsize]);
+    if (mi->use_masked_interinter) {
+      vp9_write_literal(bc, mi->mask_index, get_mask_bits(mi->sb_type));
+    }
+  }
+#endif
   }
 }
 
@@ -407,6 +485,11 @@ static void write_mb_modes_kf(const VP9_COMP *cpi, MODE_INFO **mi_8x8,
     const MB_PREDICTION_MODE A = above_block_mode(m, above_mi, 0);
     const MB_PREDICTION_MODE L = left_block_mode(m, left_mi, 0);
     write_intra_mode(bc, ym, vp9_kf_y_mode_prob[A][L]);
+#if CONFIG_FILTERINTRA
+    if (is_filter_allowed(ym) && is_filter_enabled(m->mbmi.tx_size))
+      vp9_write(bc, m->mbmi.filterbit,
+                cm->fc.filterintra_prob[m->mbmi.tx_size][ym]);
+#endif
   } else {
     int idx, idy;
     const int num_4x4_blocks_wide = num_4x4_blocks_wide_lookup[m->mbmi.sb_type];
@@ -421,11 +504,21 @@ static void write_mb_modes_kf(const VP9_COMP *cpi, MODE_INFO **mi_8x8,
         ++intra_mode_stats[A][L][bm];
 #endif
         write_intra_mode(bc, bm, vp9_kf_y_mode_prob[A][L]);
+#if CONFIG_FILTERINTRA
+        if (is_filter_allowed(bm))
+          vp9_write(bc, m->b_filter_info[i], cm->fc.filterintra_prob[0][bm]);
+#endif
       }
     }
   }
 
   write_intra_mode(bc, m->mbmi.uv_mode, vp9_kf_uv_mode_prob[ym]);
+#if CONFIG_FILTERINTRA
+  if (is_filter_allowed(m->mbmi.uv_mode) &&
+      is_filter_enabled(get_uv_tx_size(&(m->mbmi))))
+    vp9_write(bc, m->mbmi.uv_filterbit,
+          cm->fc.filterintra_prob[get_uv_tx_size(&(m->mbmi))][m->mbmi.uv_mode]);
+#endif
 }
 
 static void write_modes_b(VP9_COMP *cpi, const TileInfo *const tile,
@@ -1232,6 +1325,10 @@ static size_t write_compressed_header(VP9_COMP *cpi, uint8_t *data) {
     active_section = 1;
 #endif
 
+#if CONFIG_EXT_TX
+    vp9_cond_prob_diff_update(&header_bc, &fc->ext_tx_prob, cm->counts.ext_tx);
+#endif
+
     for (i = 0; i < INTER_MODE_CONTEXTS; ++i)
       prob_diff_update(vp9_inter_mode_tree, cm->fc.inter_mode_probs[i],
                        cm->counts.inter_mode[i], INTER_MODES, &header_bc);
@@ -1269,10 +1366,100 @@ static size_t write_compressed_header(VP9_COMP *cpi, uint8_t *data) {
       }
     }
 
-    if (cm->reference_mode != SINGLE_REFERENCE)
-      for (i = 0; i < REF_CONTEXTS; i++)
+    if (cm->reference_mode != SINGLE_REFERENCE) {
+      for (i = 0; i < REF_CONTEXTS; i++) {
         vp9_cond_prob_diff_update(&header_bc, &fc->comp_ref_prob[i],
                                   cpi->comp_ref_count[i]);
+      }
+    }
+
+#if CONFIG_MASKED_INTERINTER
+    if (cm->reference_mode != SINGLE_REFERENCE) {
+      if (!cpi->dummy_packing && cm->use_masked_interinter) {
+        cm->use_masked_interinter = 0;
+        for (i = 0; i < BLOCK_SIZES; i++) {
+          if (get_mask_bits(i) && (cpi->masked_interinter_counts[i][1] > 0)) {
+            cm->use_masked_interinter = 1;
+            break;
+          }
+        }
+      }
+      vp9_write_bit(&header_bc, cm->use_masked_interinter);
+      if (cm->use_masked_interinter) {
+        for (i = 0; i < BLOCK_SIZES; i++) {
+          if (get_mask_bits(i)) {
+            vp9_cond_prob_diff_update(&header_bc,
+                                      &fc->masked_interinter_prob[i],
+                                      cpi->masked_interinter_counts[i]);
+          }
+        }
+      } else {
+        vp9_zero(cpi->masked_interinter_counts);
+      }
+    } else {
+      if (!cpi->dummy_packing)
+        cm->use_masked_interinter = 0;
+      vp9_zero(cpi->masked_interinter_counts);
+    }
+#endif
+
+#if CONFIG_INTERINTRA
+    if (cm->reference_mode != COMPOUND_REFERENCE) {
+      if (!cpi->dummy_packing && cm->use_interintra) {
+        cm->use_interintra = 0;
+        for (i = 0; i < BLOCK_SIZES; i++) {
+          if (is_interintra_allowed(i) && (cpi->interintra_count[i][1] > 0)) {
+            cm->use_interintra = 1;
+            break;
+          }
+        }
+      }
+      vp9_write_bit(&header_bc, cm->use_interintra);
+      if (cm->use_interintra) {
+        for (i = 0; i < BLOCK_SIZES; i++) {
+          if (is_interintra_allowed(i)) {
+            vp9_cond_prob_diff_update(&header_bc,
+                                      &fc->interintra_prob[i],
+                                      cpi->interintra_count[i]);
+          }
+        }
+#if CONFIG_MASKED_INTERINTRA
+        if (!cpi->dummy_packing && cm->use_masked_interintra) {
+          cm->use_masked_interintra = 0;
+          for (i = 0; i < BLOCK_SIZES; i++) {
+            if (is_interintra_allowed(i) && get_mask_bits_interintra(i) &&
+                (cpi->masked_interintra_count[i][1] > 0)) {
+              cm->use_masked_interintra = 1;
+              break;
+            }
+          }
+        }
+        vp9_write_bit(&header_bc, cm->use_masked_interintra);
+        if (cm->use_masked_interintra) {
+          for (i = 0; i < BLOCK_SIZES; i++) {
+            if (is_interintra_allowed(i) && get_mask_bits_interintra(i))
+              vp9_cond_prob_diff_update(&header_bc,
+                                        &fc->masked_interintra_prob[i],
+                                        cpi->masked_interintra_count[i]);
+          }
+        } else {
+          vp9_zero(cpi->masked_interintra_count);
+        }
+#endif
+      } else {
+        vp9_zero(cpi->interintra_count);
+      }
+    } else {
+      if (!cpi->dummy_packing)
+        cm->use_interintra = 0;
+      vp9_zero(cpi->interintra_count);
+#if CONFIG_MASKED_INTERINTRA
+      if (!cpi->dummy_packing)
+        cm->use_masked_interintra = 0;
+      vp9_zero(cpi->masked_interintra_count);
+#endif
+    }
+#endif
 
     for (i = 0; i < BLOCK_SIZE_GROUPS; ++i)
       prob_diff_update(vp9_intra_mode_tree, cm->fc.y_mode_prob[i],

vp9/encoder/vp9_dct.c

@@ -1345,6 +1345,403 @@ void vp9_fdct32x32_c(const int16_t *input, int16_t *out, int stride) {
   }
 }
 
+#if CONFIG_EXT_TX
+static double tmp[32*32];
+static double tmp2[32*32];
+
+double dstmtx32[32*32] = {
+    0.0122669185818545, 0.0245042850823902, 0.0366826186138404,
+    0.0487725805040321, 0.0607450449758160, 0.0725711693136156,
+    0.0842224633480550, 0.0956708580912724, 0.1068887733575705,
+    0.1178491842064994, 0.1285256860483054, 0.1388925582549005,
+    0.1489248261231083, 0.1585983210409114, 0.1678897387117546,
+    0.1767766952966369, 0.1852377813387398, 0.1932526133406842,
+    0.2008018828701612, 0.2078674030756363, 0.2144321525000680,
+    0.2204803160870887, 0.2259973232808608, 0.2309698831278217,
+    0.2353860162957552, 0.2392350839330522, 0.2425078132986360,
+    0.2451963201008076, 0.2472941274911953, 0.2487961816680492,
+    0.2496988640512931, 0.1767766952966369,
+    0.0366826186138404, 0.0725711693136156, 0.1068887733575705,
+    0.1388925582549005, 0.1678897387117546, 0.1932526133406842,
+    0.2144321525000680, 0.2309698831278217, 0.2425078132986360,
+    0.2487961816680492, 0.2496988640512931, 0.2451963201008076,
+    0.2353860162957552, 0.2204803160870888, 0.2008018828701612,
+    0.1767766952966369, 0.1489248261231084, 0.1178491842064995,
+    0.0842224633480551, 0.0487725805040322, 0.0122669185818545,
+    -0.0245042850823901, -0.0607450449758160, -0.0956708580912724,
+    -0.1285256860483054, -0.1585983210409113, -0.1852377813387397,
+    -0.2078674030756363, -0.2259973232808608, -0.2392350839330522,
+    -0.2472941274911952, -0.1767766952966369,
+    0.0607450449758160, 0.1178491842064994, 0.1678897387117546,
+    0.2078674030756363, 0.2353860162957552, 0.2487961816680492,
+    0.2472941274911953, 0.2309698831278217, 0.2008018828701612,
+    0.1585983210409114, 0.1068887733575705, 0.0487725805040322,
+    -0.0122669185818544, -0.0725711693136155, -0.1285256860483054,
+    -0.1767766952966369, -0.2144321525000680, -0.2392350839330522,
+    -0.2496988640512931, -0.2451963201008076, -0.2259973232808608,
+    -0.1932526133406842, -0.1489248261231083, -0.0956708580912726,
+    -0.0366826186138406, 0.0245042850823900, 0.0842224633480549,
+    0.1388925582549005, 0.1852377813387397, 0.2204803160870887,
+    0.2425078132986360, 0.1767766952966369,
+    0.0842224633480550, 0.1585983210409114, 0.2144321525000680,
+    0.2451963201008076, 0.2472941274911953, 0.2204803160870888,
+    0.1678897387117546, 0.0956708580912725, 0.0122669185818545,
+    -0.0725711693136155, -0.1489248261231083, -0.2078674030756363,
+    -0.2425078132986360, -0.2487961816680492, -0.2259973232808608,
+    -0.1767766952966369, -0.1068887733575706, -0.0245042850823901,
+    0.0607450449758159, 0.1388925582549005, 0.2008018828701612,
+    0.2392350839330522, 0.2496988640512931, 0.2309698831278217,
+    0.1852377813387399, 0.1178491842064995, 0.0366826186138404,
+    -0.0487725805040318, -0.1285256860483053, -0.1932526133406843,
+    -0.2353860162957551, -0.1767766952966369,
+    0.1068887733575705, 0.1932526133406842, 0.2425078132986360,
+    0.2451963201008076, 0.2008018828701612, 0.1178491842064995,
+    0.0122669185818545, -0.0956708580912724, -0.1852377813387397,
+    -0.2392350839330522, -0.2472941274911952, -0.2078674030756364,
+    -0.1285256860483055, -0.0245042850823901, 0.0842224633480549,
+    0.1767766952966368, 0.2353860162957552, 0.2487961816680492,
+    0.2144321525000681, 0.1388925582549006, 0.0366826186138404,
+    -0.0725711693136157, -0.1678897387117544, -0.2309698831278216,
+    -0.2496988640512931, -0.2204803160870888, -0.1489248261231083,
+    -0.0487725805040320, 0.0607450449758156, 0.1585983210409112,
+    0.2259973232808608, 0.1767766952966369,
+    0.1285256860483054, 0.2204803160870887, 0.2496988640512931,
+    0.2078674030756363, 0.1068887733575706, -0.0245042850823900,
+    -0.1489248261231082, -0.2309698831278216, -0.2472941274911953,
+    -0.1932526133406844, -0.0842224633480552, 0.0487725805040318,
+    0.1678897387117544, 0.2392350839330522, 0.2425078132986361,
+    0.1767766952966371, 0.0607450449758163, -0.0725711693136153,
+    -0.1852377813387395, -0.2451963201008076, -0.2353860162957553,
+    -0.1585983210409117, -0.0366826186138409, 0.0956708580912720,
+    0.2008018828701609, 0.2487961816680492, 0.2259973232808611,
+    0.1388925582549010, 0.0122669185818551, -0.1178491842064989,
+    -0.2144321525000677, -0.1767766952966369,
+    0.1489248261231083, 0.2392350839330522, 0.2353860162957552,
+    0.1388925582549005, -0.0122669185818545, -0.1585983210409114,
+    -0.2425078132986360, -0.2309698831278217, -0.1285256860483055,
+    0.0245042850823902, 0.1678897387117547, 0.2451963201008076,
+    0.2259973232808608, 0.1178491842064991, -0.0366826186138407,
+    -0.1767766952966370, -0.2472941274911953, -0.2204803160870888,
+    -0.1068887733575703, 0.0487725805040322, 0.1852377813387398,
+    0.2487961816680493, 0.2144321525000679, 0.0956708580912723,
+    -0.0607450449758161, -0.1932526133406842, -0.2496988640512931,
+    -0.2078674030756359, -0.0842224633480545, 0.0725711693136160,
+    0.2008018828701614, 0.1767766952966369,
+    0.1678897387117546, 0.2487961816680492, 0.2008018828701613,
+    0.0487725805040322, -0.1285256860483054, -0.2392350839330522,
+    -0.2259973232808609, -0.0956708580912726, 0.0842224633480549,
+    0.2204803160870887, 0.2425078132986361, 0.1388925582549010,
+    -0.0366826186138402, -0.1932526133406840, -0.2496988640512931,
+    -0.1767766952966371, -0.0122669185818550, 0.1585983210409112,
+    0.2472941274911952, 0.2078674030756364, 0.0607450449758163,
+    -0.1178491842064989, -0.2353860162957550, -0.2309698831278220,
+    -0.1068887733575707, 0.0725711693136152, 0.2144321525000677,
+    0.2451963201008078, 0.1489248261231091, -0.0245042850823899,
+    -0.1852377813387394, -0.1767766952966369,
+    0.1852377813387398, 0.2487961816680492, 0.1489248261231083,
+    -0.0487725805040321, -0.2144321525000680, -0.2392350839330522,
+    -0.1068887733575704, 0.0956708580912725, 0.2353860162957552,
+    0.2204803160870888, 0.0607450449758159, -0.1388925582549008,
+    -0.2472941274911953, -0.1932526133406841, -0.0122669185818546,
+    0.1767766952966369, 0.2496988640512931, 0.1585983210409114,
+    -0.0366826186138406, -0.2078674030756363, -0.2425078132986359,
+    -0.1178491842064992, 0.0842224633480550, 0.2309698831278219,
+    0.2259973232808607, 0.0725711693136156, -0.1285256860483052,
+    -0.2451963201008077, -0.2008018828701612, -0.0245042850823903,
+    0.1678897387117549, 0.1767766952966369,
+    0.2008018828701612, 0.2392350839330522, 0.0842224633480551,
+    -0.1388925582549005, -0.2496988640512931, -0.1585983210409115,
+    0.0607450449758157, 0.2309698831278216, 0.2144321525000681,
+    0.0245042850823902, -0.1852377813387395, -0.2451963201008077,
+    -0.1068887733575707, 0.1178491842064989, 0.2472941274911952,
+    0.1767766952966371, -0.0366826186138402, -0.2204803160870887,
+    -0.2259973232808611, -0.0487725805040321, 0.1678897387117543,
+    0.2487961816680493, 0.1285256860483056, -0.0956708580912719,
+    -0.2425078132986358, -0.1932526133406845, 0.0122669185818538,
+    0.2078674030756357, 0.2353860162957554, 0.0725711693136165,
+    -0.1489248261231080, -0.1767766952966369,
+    0.2144321525000680, 0.2204803160870888, 0.0122669185818545,
+    -0.2078674030756363, -0.2259973232808608, -0.0245042850823901,
+    0.2008018828701612, 0.2309698831278217, 0.0366826186138404,
+    -0.1932526133406843, -0.2353860162957552, -0.0487725805040320,
+    0.1852377813387398, 0.2392350839330522, 0.0607450449758159,
+    -0.1767766952966369, -0.2425078132986359, -0.0725711693136155,
+    0.1678897387117543, 0.2451963201008076, 0.0842224633480545,
+    -0.1585983210409114, -0.2472941274911953, -0.0956708580912724,
+    0.1489248261231088, 0.2487961816680492, 0.1068887733575708,
+    -0.1388925582549007, -0.2496988640512931, -0.1178491842064993,
+    0.1285256860483052, 0.1767766952966369,
+    0.2259973232808608, 0.1932526133406843, -0.0607450449758160,
+    -0.2451963201008076, -0.1489248261231085, 0.1178491842064994,
+    0.2496988640512931, 0.0956708580912726, -0.1678897387117544,
+    -0.2392350839330523, -0.0366826186138409, 0.2078674030756363,
+    0.2144321525000681, -0.0245042850823899, -0.2353860162957550,
+    -0.1767766952966372, 0.0842224633480550, 0.2487961816680492,
+    0.1285256860483056, -0.1388925582549000, -0.2472941274911953,
+    -0.0725711693136164, 0.1852377813387394, 0.2309698831278217,
+    0.0122669185818552, -0.2204803160870886, -0.2008018828701617,
+    0.0487725805040316, 0.2425078132986357, 0.1585983210409125,
+    -0.1068887733575703, -0.1767766952966369,
+    0.2353860162957552, 0.1585983210409114, -0.1285256860483054,
+    -0.2451963201008076, -0.0366826186138404, 0.2204803160870887,
+    0.1852377813387396, -0.0956708580912725, -0.2496988640512931,
+    -0.0725711693136155, 0.2008018828701612, 0.2078674030756364,
+    -0.0607450449758161, -0.2487961816680493, -0.1068887733575707,
+    0.1767766952966369, 0.2259973232808607, -0.0245042850823899,
+    -0.2425078132986360, -0.1388925582549004, 0.1489248261231081,
+    0.2392350839330522, 0.0122669185818543, -0.2309698831278215,
+    -0.1678897387117547, 0.1178491842064995, 0.2472941274911952,
+    0.0487725805040314, -0.2144321525000676, -0.1932526133406846,
+    0.0842224633480548, 0.1767766952966369,
+    0.2425078132986360, 0.1178491842064995, -0.1852377813387397,
+    -0.2078674030756364, 0.0842224633480549, 0.2487961816680492,
+    0.0366826186138408, -0.2309698831278216, -0.1489248261231083,
+    0.1585983210409112, 0.2259973232808611, -0.0487725805040317,
+    -0.2496988640512931, -0.0725711693136164, 0.2144321525000677,
+    0.1767766952966372, -0.1285256860483052, -0.2392350839330522,
+    0.0122669185818538, 0.2451963201008075, 0.1068887733575708,
+    -0.1932526133406836, -0.2008018828701617, 0.0956708580912718,
+    0.2472941274911954, 0.0245042850823904, -0.2353860162957549,
+    -0.1388925582549019, 0.1678897387117542, 0.2204803160870893,
+    -0.0607450449758158, -0.1767766952966369,
+    0.2472941274911953, 0.0725711693136156, -0.2259973232808608,
+    -0.1388925582549005, 0.1852377813387399, 0.1932526133406844,
+    -0.1285256860483053, -0.2309698831278217, 0.0607450449758161,
+    0.2487961816680492, 0.0122669185818546, -0.2451963201008075,
+    -0.0842224633480553, 0.2204803160870887, 0.1489248261231084,
+    -0.1767766952966369, -0.2008018828701612, 0.1178491842064996,
+    0.2353860162957551, -0.0487725805040325, -0.2496988640512931,
+    -0.0245042850823904, 0.2425078132986362, 0.0956708580912733,
+    -0.2144321525000676, -0.1585983210409119, 0.1678897387117542,
+    0.2078674030756366, -0.1068887733575702, -0.2392350839330523,
+    0.0366826186138403, 0.1767766952966369,
+    0.2496988640512931, 0.0245042850823902, -0.2472941274911952,
+    -0.0487725805040322, 0.2425078132986360, 0.0725711693136159,
+    -0.2353860162957551, -0.0956708580912727, 0.2259973232808608,
+    0.1178491842064996, -0.2144321525000677, -0.1388925582549010,
+    0.2008018828701609, 0.1585983210409118, -0.1852377813387394,
+    -0.1767766952966372, 0.1678897387117543, 0.1932526133406845,
+    -0.1489248261231080, -0.2078674030756365, 0.1285256860483052,
+    0.2204803160870893, -0.1068887733575703, -0.2309698831278221,
+    0.0842224633480548, 0.2392350839330525, -0.0607450449758158,
+    -0.2451963201008078, 0.0366826186138403, 0.2487961816680493,
+    -0.0122669185818544, -0.1767766952966369,
+    0.2496988640512931, -0.0245042850823901, -0.2472941274911952,
+    0.0487725805040321, 0.2425078132986360, -0.0725711693136157,
+    -0.2353860162957552, 0.0956708580912724, 0.2259973232808609,
+    -0.1178491842064997, -0.2144321525000679, 0.1388925582549007,
+    0.2008018828701612, -0.1585983210409114, -0.1852377813387397,
+    0.1767766952966369, 0.1678897387117547, -0.1932526133406841,
+    -0.1489248261231085, 0.2078674030756367, 0.1285256860483049,
+    -0.2204803160870890, -0.1068887733575701, 0.2309698831278219,
+    0.0842224633480547, -0.2392350839330523, -0.0607450449758158,
+    0.2451963201008076, 0.0366826186138403, -0.2487961816680492,
+    -0.0122669185818545, 0.1767766952966369,
+    0.2472941274911953, -0.0725711693136155, -0.2259973232808608,
+    0.1388925582549005, 0.1852377813387399, -0.1932526133406843,
+    -0.1285256860483055, 0.2309698831278216, 0.0607450449758163,
+    -0.2487961816680492, 0.0122669185818538, 0.2451963201008076,
+    -0.0842224633480550, -0.2204803160870888, 0.1489248261231081,
+    0.1767766952966372, -0.2008018828701608, -0.1178491842065001,
+    0.2353860162957552, 0.0487725805040331, -0.2496988640512931,
+    0.0245042850823888, 0.2425078132986361, -0.0956708580912726,
+    -0.2144321525000685, 0.1585983210409113, 0.1678897387117555,
+    -0.2078674030756361, -0.1068887733575718, 0.2392350839330520,
+    0.0366826186138421, -0.1767766952966369,
+    0.2425078132986360, -0.1178491842064994, -0.1852377813387398,
+    0.2078674030756363, 0.0842224633480552, -0.2487961816680492,
+    0.0366826186138406, 0.2309698831278217, -0.1489248261231081,
+    -0.1585983210409118, 0.2259973232808609, 0.0487725805040321,
+    -0.2496988640512931, 0.0725711693136160, 0.2144321525000679,
+    -0.1767766952966368, -0.1285256860483057, 0.2392350839330521,
+    0.0122669185818552, -0.2451963201008078, 0.1068887733575710,
+    0.1932526133406840, -0.2008018828701613, -0.0956708580912725,
+    0.2472941274911952, -0.0245042850823896, -0.2353860162957555,
+    0.1388925582549012, 0.1678897387117542, -0.2204803160870889,
+    -0.0607450449758159, 0.1767766952966369,
+    0.2353860162957552, -0.1585983210409113, -0.1285256860483055,
+    0.2451963201008076, -0.0366826186138402, -0.2204803160870888,
+    0.1852377813387395, 0.0956708580912727, -0.2496988640512931,
+    0.0725711693136152, 0.2008018828701617, -0.2078674030756362,
+    -0.0607450449758164, 0.2487961816680493, -0.1068887733575703,
+    -0.1767766952966373, 0.2259973232808605, 0.0245042850823904,
+    -0.2425078132986364, 0.1388925582548998, 0.1489248261231086,
+    -0.2392350839330518, 0.0122669185818536, 0.2309698831278218,
+    -0.1678897387117535, -0.1178491842065003, 0.2472941274911952,
+    -0.0487725805040305, -0.2144321525000685, 0.1932526133406840,
+    0.0842224633480565, -0.1767766952966369,
+    0.2259973232808609, -0.1932526133406842, -0.0607450449758163,
+    0.2451963201008077, -0.1489248261231082, -0.1178491842064999,
+    0.2496988640512931, -0.0956708580912720, -0.1678897387117552,
+    0.2392350839330521, -0.0366826186138397, -0.2078674030756370,
+    0.2144321525000677, 0.0245042850823912, -0.2353860162957554,
+    0.1767766952966362, 0.0842224633480563, -0.2487961816680494,
+    0.1285256860483051, 0.1388925582549012, -0.2472941274911951,
+    0.0725711693136141, 0.1852377813387411, -0.2309698831278208,
+    0.0122669185818535, 0.2204803160870894, -0.2008018828701602,
+    -0.0487725805040342, 0.2425078132986366, -0.1585983210409105,
+    -0.1068887733575719, 0.1767766952966369,
+    0.2144321525000680, -0.2204803160870887, 0.0122669185818544,
+    0.2078674030756364, -0.2259973232808608, 0.0245042850823899,
+    0.2008018828701614, -0.2309698831278216, 0.0366826186138406,
+    0.1932526133406845, -0.2353860162957549, 0.0487725805040317,
+    0.1852377813387398, -0.2392350839330521, 0.0607450449758150,
+    0.1767766952966373, -0.2425078132986360, 0.0725711693136158,
+    0.1678897387117554, -0.2451963201008075, 0.0842224633480547,
+    0.1585983210409126, -0.2472941274911951, 0.0956708580912717,
+    0.1489248261231087, -0.2487961816680492, 0.1068887733575693,
+    0.1388925582549013, -0.2496988640512931, 0.1178491842064977,
+    0.1285256860483052, -0.1767766952966369,
+    0.2008018828701612, -0.2392350839330522, 0.0842224633480551,
+    0.1388925582549006, -0.2496988640512931, 0.1585983210409115,
+    0.0607450449758159, -0.2309698831278217, 0.2144321525000682,
+    -0.0245042850823899, -0.1852377813387397, 0.2451963201008077,
+    -0.1068887733575703, -0.1178491842064993, 0.2472941274911952,
+    -0.1767766952966368, -0.0366826186138411, 0.2204803160870885,
+    -0.2259973232808608, 0.0487725805040315, 0.1678897387117541,
+    -0.2487961816680492, 0.1285256860483050, 0.0956708580912718,
+    -0.2425078132986360, 0.1932526133406840, 0.0122669185818555,
+    -0.2078674030756362, 0.2353860162957551, -0.0725711693136164,
+    -0.1489248261231095, 0.1767766952966369,
+    0.1852377813387397, -0.2487961816680492, 0.1489248261231084,
+    0.0487725805040320, -0.2144321525000679, 0.2392350839330523,
+    -0.1068887733575704, -0.0956708580912723, 0.2353860162957551,
+    -0.2204803160870891, 0.0607450449758160, 0.1388925582549004,
+    -0.2472941274911952, 0.1932526133406841, -0.0122669185818546,
+    -0.1767766952966367, 0.2496988640512931, -0.1585983210409120,
+    -0.0366826186138394, 0.2078674030756356, -0.2425078132986359,
+    0.1178491842064994, 0.0842224633480548, -0.2309698831278215,
+    0.2259973232808611, -0.0725711693136165, -0.1285256860483044,
+    0.2451963201008077, -0.2008018828701612, 0.0245042850823903,
+    0.1678897387117543, -0.1767766952966369,
+    0.1678897387117546, -0.2487961816680492, 0.2008018828701611,
+    -0.0487725805040318, -0.1285256860483055, 0.2392350839330523,
+    -0.2259973232808606, 0.0956708580912720, 0.0842224633480554,
+    -0.2204803160870888, 0.2425078132986358, -0.1388925582548999,
+    -0.0366826186138411, 0.1932526133406851, -0.2496988640512931,
+    0.1767766952966361, -0.0122669185818527, -0.1585983210409119,
+    0.2472941274911954, -0.2078674030756361, 0.0607450449758149,
+    0.1178491842065011, -0.2353860162957555, 0.2309698831278211,
+    -0.1068887733575685, -0.0725711693136168, 0.2144321525000690,
+    -0.2451963201008071, 0.1489248261231056, 0.0245042850823908,
+    -0.1852377813387407, 0.1767766952966369,
+    0.1489248261231084, -0.2392350839330522, 0.2353860162957551,
+    -0.1388925582549004, -0.0122669185818546, 0.1585983210409117,
+    -0.2425078132986361, 0.2309698831278216, -0.1285256860483052,
+    -0.0245042850823903, 0.1678897387117547, -0.2451963201008078,
+    0.2259973232808605, -0.1178491842064987, -0.0366826186138411,
+    0.1767766952966373, -0.2472941274911953, 0.2204803160870885,
+    -0.1068887733575702, -0.0487725805040324, 0.1852377813387399,
+    -0.2487961816680492, 0.2144321525000680, -0.0956708580912708,
+    -0.0607450449758177, 0.1932526133406853, -0.2496988640512931,
+    0.2078674030756355, -0.0842224633480553, -0.0725711693136169,
+    0.2008018828701609, -0.1767766952966369,
+    0.1285256860483054, -0.2204803160870888, 0.2496988640512931,
+    -0.2078674030756363, 0.1068887733575705, 0.0245042850823902,
+    -0.1489248261231084, 0.2309698831278217, -0.2472941274911952,
+    0.1932526133406842, -0.0842224633480549, -0.0487725805040322,
+    0.1678897387117541, -0.2392350839330523, 0.2425078132986357,
+    -0.1767766952966368, 0.0607450449758166, 0.0725711693136158,
+    -0.1852377813387405, 0.2451963201008077, -0.2353860162957554,
+    0.1585983210409112, -0.0366826186138393, -0.0956708580912727,
+    0.2008018828701609, -0.2487961816680491, 0.2259973232808603,
+    -0.1388925582549003, 0.0122669185818551, 0.1178491842065013,
+    -0.2144321525000687, 0.1767766952966369,
+    0.1068887733575705, -0.1932526133406842, 0.2425078132986360,
+    -0.2451963201008076, 0.2008018828701612, -0.1178491842064997,
+    0.0122669185818547, 0.0956708580912723, -0.1852377813387397,
+    0.2392350839330522, -0.2472941274911952, 0.2078674030756367,
+    -0.1285256860483059, 0.0245042850823906, 0.0842224633480547,
+    -0.1767766952966367, 0.2353860162957551, -0.2487961816680492,
+    0.2144321525000680, -0.1388925582549004, 0.0366826186138402,
+    0.0725711693136159, -0.1678897387117549, 0.2309698831278212,
+    -0.2496988640512932, 0.2204803160870893, -0.1489248261231091,
+    0.0487725805040329, 0.0607450449758152, -0.1585983210409108,
+    0.2259973232808606, -0.1767766952966369,
+    0.0842224633480551, -0.1585983210409115, 0.2144321525000681,
+    -0.2451963201008077, 0.2472941274911952, -0.2204803160870887,
+    0.1678897387117543, -0.0956708580912719, 0.0122669185818538,
+    0.0725711693136165, -0.1489248261231092, 0.2078674030756365,
+    -0.2425078132986364, 0.2487961816680491, -0.2259973232808601,
+    0.1767766952966361, -0.1068887733575702, 0.0245042850823887,
+    0.0607450449758167, -0.1388925582549021, 0.2008018828701619,
+    -0.2392350839330528, 0.2496988640512929, -0.2309698831278214,
+    0.1852377813387385, -0.1178491842064969, 0.0366826186138400,
+    0.0487725805040335, -0.1285256860483076, 0.1932526133406866,
+    -0.2353860162957556, 0.1767766952966369,
+    0.0607450449758160, -0.1178491842064995, 0.1678897387117545,
+    -0.2078674030756364, 0.2353860162957553, -0.2487961816680492,
+    0.2472941274911952, -0.2309698831278216, 0.2008018828701609,
+    -0.1585983210409107, 0.1068887733575703, -0.0487725805040324,
+    -0.0122669185818544, 0.0725711693136157, -0.1285256860483058,
+    0.1767766952966374, -0.2144321525000685, 0.2392350839330526,
+    -0.2496988640512932, 0.2451963201008073, -0.2259973232808608,
+    0.1932526133406839, -0.1489248261231077, 0.0956708580912732,
+    -0.0366826186138392, -0.0245042850823899, 0.0842224633480567,
+    -0.1388925582549008, 0.1852377813387413, -0.2204803160870892,
+    0.2425078132986367, -0.1767766952966369,
+    0.0366826186138405, -0.0725711693136156, 0.1068887733575707,
+    -0.1388925582549006, 0.1678897387117546, -0.1932526133406844,
+    0.2144321525000679, -0.2309698831278217, 0.2425078132986361,
+    -0.2487961816680492, 0.2496988640512931, -0.2451963201008075,
+    0.2353860162957552, -0.2204803160870890, 0.2008018828701608,
+    -0.1767766952966367, 0.1489248261231086, -0.1178491842064986,
+    0.0842224633480546, -0.0487725805040322, 0.0122669185818534,
+    0.0245042850823889, -0.0607450449758177, 0.0956708580912736,
+    -0.1285256860483061, 0.1585983210409115, -0.1852377813387395,
+    0.2078674030756358, -0.2259973232808617, 0.2392350839330527,
+    -0.2472941274911954, 0.1767766952966369,
+    0.0122669185818545, -0.0245042850823901, 0.0366826186138404,
+    -0.0487725805040320, 0.0607450449758159, -0.0725711693136155,
+    0.0842224633480545, -0.0956708580912724, 0.1068887733575708,
+    -0.1178491842064993, 0.1285256860483049, -0.1388925582549004,
+    0.1489248261231086, -0.1585983210409106, 0.1678897387117541,
+    -0.1767766952966368, 0.1852377813387399, -0.1932526133406847,
+    0.2008018828701608, -0.2078674030756362, 0.2144321525000672,
+    -0.2204803160870882, 0.2259973232808605, -0.2309698831278216,
+    0.2353860162957553, -0.2392350839330524, 0.2425078132986362,
+    -0.2451963201008072, 0.2472941274911950, -0.2487961816680491,
+    0.2496988640512930, -0.1767766952966369,
+};
+
+void vp9_fdst(const int16_t *input, int16_t *output, int stride, int size) {
+  int i, j, k;
+  double *basis;
+  int factor = (size == 32) ? 4 : 8;
+
+  switch (size) {
+    case 32:
+      basis = dstmtx32;
+      break;
+    default:
+      assert(0);
+      break;
+  }
+
+  for (i = 0; i < size; i++) {
+    for (j = 0; j < size; j++) {
+      tmp[i*size+j] = 0;
+      for (k = 0; k < size; k++) {
+        tmp[i*size+j] += input[i*stride+k] * basis[k*size+j];  // row
+      }
+    }
+  }
+
+  for (i = 0; i < size; i++) {
+    for (j = 0; j < size; j++) {
+      tmp2[i*size+j] = 0;
+      for (k = 0; k < size; k++) {
+        tmp2[i*size+j] += basis[k*size+i] * tmp[k*size+j];  // col
+      }
+      if (tmp2[i*size+j] >= 0)
+        output[i*size+j] = (int16_t)(tmp2[i*size+j] * factor + 0.5);
+      else
+        output[i*size+j] = (int16_t)(tmp2[i*size+j] * factor - 0.5);
+    }
+  }
+}
+#endif
+
 // Note that although we use dct_32_round in dct32_1d computation flow,
 // this 2d fdct32x32 for rate-distortion optimization loop is operating
 // within 16 bits precision.

vp9/encoder/vp9_dct.h

@@ -12,6 +12,10 @@
 #ifndef VP9_ENCODER_VP9_DCT_H_
 #define VP9_ENCODER_VP9_DCT_H_
 
+#if CONFIG_EXT_TX
+int block_index;
+#endif
+
 void vp9_fht4x4(TX_TYPE tx_type, const int16_t *input, int16_t *output,
                 int stride);
 

vp9/encoder/vp9_encodeframe.c

@@ -526,6 +526,40 @@ static void update_state(VP9_COMP *cpi, PICK_MODE_CONTEXT *ctx,
       vp9_update_mv_count(cpi, x, best_mv);
     }
 
+#if CONFIG_MASKED_INTERINTER
+    if (cm->use_masked_interinter &&
+        cm->reference_mode != SINGLE_REFERENCE &&
+        is_inter_mode(mbmi->mode) &&
+        get_mask_bits(bsize) &&
+        mbmi->ref_frame[1] > INTRA_FRAME) {
+      ++cpi->masked_interinter_counts[bsize][mbmi->use_masked_interinter];
+    }
+#endif
+
+#if CONFIG_INTERINTRA
+    if (cm->use_interintra &&
+        is_interintra_allowed(mbmi->sb_type) &&
+        is_inter_mode(mbmi->mode) &&
+        (mbmi->ref_frame[1] <= INTRA_FRAME)) {
+      if (mbmi->ref_frame[1] == INTRA_FRAME) {
+        ++cpi->y_mode_count[size_group_lookup[mbmi->sb_type]]
+                           [mbmi->interintra_mode];
+#if SEPARATE_INTERINTRA_UV
+        ++cpi->y_uv_mode_count[mbmi->interintra_mode][mbmi->interintra_uv_mode];
+#endif
+        ++cpi->interintra_count[mbmi->sb_type][1];
+#if CONFIG_MASKED_INTERINTRA
+        if (cm->use_masked_interintra &&
+            get_mask_bits_interintra(mbmi->sb_type))
+          ++cpi->masked_interintra_count[mbmi->sb_type]
+                                        [mbmi->use_masked_interintra];
+#endif
+      } else {
+        ++cpi->interintra_count[mbmi->sb_type][0];
+      }
+    }
+#endif
+
     if (cm->mcomp_filter_type == SWITCHABLE && is_inter_mode(mbmi->mode)) {
       const int ctx = vp9_get_pred_context_switchable_interp(xd);
       ++cm->counts.switchable_interp[ctx][mbmi->interp_filter];
@@ -2005,6 +2039,10 @@ static void init_encode_frame_mb_context(VP9_COMP *cpi) {
 
   xd->mi_8x8[0]->mbmi.mode = DC_PRED;
   xd->mi_8x8[0]->mbmi.uv_mode = DC_PRED;
+#if CONFIG_FILTERINTRA
+  xd->mi_8x8[0]->mbmi.filterbit = 0;
+  xd->mi_8x8[0]->mbmi.uv_filterbit = 0;
+#endif
 
   vp9_zero(cpi->y_mode_count);
   vp9_zero(cpi->y_uv_mode_count);
@@ -2017,6 +2055,25 @@ static void init_encode_frame_mb_context(VP9_COMP *cpi) {
   vp9_zero(cm->counts.tx);
   vp9_zero(cm->counts.mbskip);
 
+#if CONFIG_FILTERINTRA
+  vp9_zero(cm->counts.filterintra);
+#endif
+#if CONFIG_MASKED_INTERINTER
+  vp9_zero(cpi->masked_interinter_counts);
+  vp9_zero(cpi->masked_interinter_select_counts);
+#endif
+#if CONFIG_INTERINTRA
+  vp9_zero(cpi->interintra_count);
+  vp9_zero(cpi->interintra_select_count);
+#if CONFIG_MASKED_INTERINTRA
+  vp9_zero(cpi->masked_interintra_count);
+  vp9_zero(cpi->masked_interintra_select_count);
+#endif
+#endif
+#if CONFIG_EXT_TX
+  vp9_zero(cm->counts.ext_tx);
+#endif
+
   // Note: this memset assumes above_context[0], [1] and [2]
   // are allocated as part of the same buffer.
   vpx_memset(cpi->above_context[0], 0,
@@ -2069,6 +2126,9 @@ static void encode_frame_internal(VP9_COMP *cpi) {
 #endif
 
   vp9_zero(cm->counts.switchable_interp);
+#if CONFIG_EXT_TX
+  vp9_zero(cm->counts.ext_tx);
+#endif
   vp9_zero(cpi->tx_stepdown_count);
 
   xd->mi_8x8 = cm->mi_grid_visible;
@@ -2488,8 +2548,19 @@ static void sum_intra_stats(VP9_COMP *cpi, const MODE_INFO *mi) {
   const MB_PREDICTION_MODE y_mode = mi->mbmi.mode;
   const MB_PREDICTION_MODE uv_mode = mi->mbmi.uv_mode;
   const BLOCK_SIZE bsize = mi->mbmi.sb_type;
+#if CONFIG_FILTERINTRA
+  const int uv_fbit = mi->mbmi.uv_filterbit;
+  int fbit = mi->mbmi.filterbit;
+#endif
+
 
   ++cpi->y_uv_mode_count[y_mode][uv_mode];
+#if CONFIG_FILTERINTRA
+  if (is_filter_allowed(uv_mode) &&
+      is_filter_enabled(get_uv_tx_size(&(mi->mbmi))))
+    ++cpi->common.counts.filterintra[get_uv_tx_size(&(mi->mbmi))]
+                                    [uv_mode][uv_fbit];
+#endif
 
   if (bsize < BLOCK_8X8) {
     int idx, idy;
@@ -2497,9 +2568,24 @@ static void sum_intra_stats(VP9_COMP *cpi, const MODE_INFO *mi) {
     const int num_4x4_blocks_high = num_4x4_blocks_high_lookup[bsize];
     for (idy = 0; idy < 2; idy += num_4x4_blocks_high)
       for (idx = 0; idx < 2; idx += num_4x4_blocks_wide)
+#if CONFIG_FILTERINTRA
+      {
+#endif
         ++cpi->y_mode_count[0][mi->bmi[idy * 2 + idx].as_mode];
+#if CONFIG_FILTERINTRA
+        if (is_filter_allowed(mi->bmi[idy * 2 + idx].as_mode)) {
+          fbit = mi->b_filter_info[idy * 2 + idx];
+          ++cpi->common.counts.filterintra[0][mi->bmi[idy * 2 + idx].as_mode]
+                                          [fbit];
+        }
+      }
+#endif
   } else {
     ++cpi->y_mode_count[size_group_lookup[bsize]][y_mode];
+#if CONFIG_FILTERINTRA
+    if (is_filter_allowed(y_mode) && is_filter_enabled(mi->mbmi.tx_size))
+      ++cpi->common.counts.filterintra[mi->mbmi.tx_size][y_mode][fbit];
+#endif
   }
 }
 
@@ -2645,5 +2731,19 @@ static void encode_superblock(VP9_COMP *cpi, TOKENEXTRA **t, int output_enabled,
           if (mi_col + x < cm->mi_cols && mi_row + y < cm->mi_rows)
             mi_8x8[mis * y + x]->mbmi.tx_size = tx_size;
     }
+
+#if CONFIG_EXT_TX
+#if CONFIG_EXT_TX_DST32
+    if (is_inter_block(mbmi) &&
+            !mbmi->skip_coeff &&
+            !vp9_segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
+#else
+    if (is_inter_block(mbmi) && mbmi->tx_size <= TX_16X16 &&
+        !mbmi->skip_coeff &&
+        !vp9_segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
+#endif
+      ++cm->counts.ext_tx[mbmi->ext_txfrm];
+    }
+#endif
   }
 }

vp9/encoder/vp9_encodemb.c

@@ -25,6 +25,8 @@
 #include "vp9/encoder/vp9_rdopt.h"
 #include "vp9/encoder/vp9_tokenize.h"
 
+
+
 void vp9_setup_interp_filters(MACROBLOCKD *xd,
                               INTERPOLATION_TYPE mcomp_filter_type,
                               VP9_COMMON *cm) {
@@ -359,6 +361,7 @@ static void optimize_init_b(int plane, BLOCK_SIZE bsize,
                            num_4x4_w, num_4x4_h);
 }
 
+#if !CONFIG_EXT_TX
 void vp9_xform_quant(int plane, int block, BLOCK_SIZE plane_bsize,
                      TX_SIZE tx_size, void *arg) {
   struct encode_b_args* const args = arg;
@@ -417,7 +420,161 @@ void vp9_xform_quant(int plane, int block, BLOCK_SIZE plane_bsize,
       assert(0);
   }
 }
+#else
+void vp9_xform_quant(int plane, int block, BLOCK_SIZE plane_bsize,
+                     TX_SIZE tx_size, void *arg) {
+  struct encode_b_args* const args = arg;
+  MACROBLOCK* const x = args->x;
+  MACROBLOCKD* const xd = &x->e_mbd;
+  struct macroblock_plane *const p = &x->plane[plane];
+  struct macroblockd_plane *const pd = &xd->plane[plane];
+  int16_t *coeff = BLOCK_OFFSET(p->coeff, block);
+  int16_t *qcoeff = BLOCK_OFFSET(p->qcoeff, block);
+  int16_t *dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
+  const scan_order *scan_order;
+  uint16_t *eob = &p->eobs[block];
+  const int diff_stride = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
+  int i, j;
+  int16_t *src_diff;
+  MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
+  txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &i, &j);
+  src_diff = &p->src_diff[4 * (j * diff_stride + i)];
 
+  if (plane == 0) {
+    switch (mbmi->ext_txfrm) {
+      case NORM:
+        switch (tx_size) {
+          case TX_32X32:
+            scan_order = &vp9_default_scan_orders[TX_32X32];
+            if (x->use_lp32x32fdct)
+              vp9_fdct32x32_rd(src_diff, coeff, diff_stride);
+            else
+              vp9_fdct32x32(src_diff, coeff, diff_stride);
+            vp9_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round,
+                                 p->quant, p->quant_shift, qcoeff, dqcoeff,
+                                 pd->dequant, p->zbin_extra, eob,
+                                 scan_order->scan, scan_order->iscan);
+            break;
+          case TX_16X16:
+            scan_order = &vp9_default_scan_orders[TX_16X16];
+            vp9_fdct16x16(src_diff, coeff, diff_stride);
+            vp9_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
+                           p->quant, p->quant_shift, qcoeff, dqcoeff,
+                           pd->dequant, p->zbin_extra, eob,
+                           scan_order->scan, scan_order->iscan);
+            break;
+          case TX_8X8:
+            scan_order = &vp9_default_scan_orders[TX_8X8];
+            vp9_fdct8x8(src_diff, coeff, diff_stride);
+            vp9_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
+                           p->quant, p->quant_shift, qcoeff, dqcoeff,
+                           pd->dequant, p->zbin_extra, eob,
+                           scan_order->scan, scan_order->iscan);
+            break;
+          case TX_4X4:
+            scan_order = &vp9_default_scan_orders[TX_4X4];
+            x->fwd_txm4x4(src_diff, coeff, diff_stride);
+            vp9_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
+                           p->quant, p->quant_shift, qcoeff, dqcoeff,
+                           pd->dequant, p->zbin_extra, eob,
+                           scan_order->scan, scan_order->iscan);
+            break;
+          default:
+            assert(0);
+        }
+      break;
+      case ALT:
+        switch (tx_size) {
+          case TX_32X32:
+            scan_order = &vp9_default_scan_orders[TX_32X32];
+#if CONFIG_EXT_TX_DST32
+            vp9_fdst(src_diff, coeff, diff_stride, 32);
+#else
+            if (x->use_lp32x32fdct)
+              vp9_fdct32x32_rd(src_diff, coeff, diff_stride);
+            else
+              vp9_fdct32x32(src_diff, coeff, diff_stride);
+#endif
+            vp9_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round,
+                                 p->quant, p->quant_shift, qcoeff, dqcoeff,
+                                 pd->dequant, p->zbin_extra, eob,
+                                 scan_order->scan, scan_order->iscan);
+            break;
+          case TX_16X16:
+            scan_order = &vp9_scan_orders[TX_16X16][ADST_ADST];
+            vp9_fht16x16(ADST_ADST, src_diff, coeff, diff_stride);
+            vp9_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
+                           p->quant, p->quant_shift, qcoeff, dqcoeff,
+                           pd->dequant, p->zbin_extra, eob,
+                           scan_order->scan, scan_order->iscan);
+            break;
+          case TX_8X8:
+            scan_order = &vp9_scan_orders[TX_8X8][ADST_ADST];
+            vp9_fht8x8(ADST_ADST, src_diff, coeff, diff_stride);
+            vp9_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
+                           p->quant, p->quant_shift, qcoeff, dqcoeff,
+                           pd->dequant, p->zbin_extra, eob,
+                           scan_order->scan, scan_order->iscan);
+            break;
+          case TX_4X4:
+            scan_order = &vp9_scan_orders[TX_4X4][ADST_ADST];
+            vp9_fht4x4(ADST_ADST, src_diff, coeff, diff_stride);
+            vp9_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
+                           p->quant, p->quant_shift, qcoeff, dqcoeff,
+                           pd->dequant, p->zbin_extra, eob,
+                           scan_order->scan, scan_order->iscan);
+            break;
+          default:
+            assert(0);
+        }
+        break;
+      default:
+        assert(0);
+    }
+  } else {
+    switch (tx_size) {
+      case TX_32X32:
+        scan_order = &vp9_default_scan_orders[TX_32X32];
+        if (x->use_lp32x32fdct)
+          vp9_fdct32x32_rd(src_diff, coeff, diff_stride);
+        else
+          vp9_fdct32x32(src_diff, coeff, diff_stride);
+        vp9_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round,
+                             p->quant, p->quant_shift, qcoeff, dqcoeff,
+                             pd->dequant, p->zbin_extra, eob, scan_order->scan,
+                             scan_order->iscan);
+        break;
+      case TX_16X16:
+        scan_order = &vp9_default_scan_orders[TX_16X16];
+        vp9_fdct16x16(src_diff, coeff, diff_stride);
+        vp9_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
+                       p->quant, p->quant_shift, qcoeff, dqcoeff,
+                       pd->dequant, p->zbin_extra, eob,
+                       scan_order->scan, scan_order->iscan);
+        break;
+      case TX_8X8:
+        scan_order = &vp9_default_scan_orders[TX_8X8];
+        vp9_fdct8x8(src_diff, coeff, diff_stride);
+        vp9_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
+                       p->quant, p->quant_shift, qcoeff, dqcoeff,
+                       pd->dequant, p->zbin_extra, eob,
+                       scan_order->scan, scan_order->iscan);
+        break;
+      case TX_4X4:
+        scan_order = &vp9_default_scan_orders[TX_4X4];
+        x->fwd_txm4x4(src_diff, coeff, diff_stride);
+        vp9_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
+                       p->quant, p->quant_shift, qcoeff, dqcoeff,
+                       pd->dequant, p->zbin_extra, eob,
+                       scan_order->scan, scan_order->iscan);
+        break;
+      default:
+        assert(0);
+    }
+  }
+}
+#endif
+#if !CONFIG_EXT_TX
 static void encode_block(int plane, int block, BLOCK_SIZE plane_bsize,
                          TX_SIZE tx_size, void *arg) {
   struct encode_b_args *const args = arg;
@@ -474,6 +631,115 @@ static void encode_block(int plane, int block, BLOCK_SIZE plane_bsize,
       assert(0 && "Invalid transform size");
   }
 }
+#else
+static void encode_block(int plane, int block, BLOCK_SIZE plane_bsize,
+                         TX_SIZE tx_size, void *arg) {
+  struct encode_b_args *const args = arg;
+  MACROBLOCK *const x = args->x;
+  MACROBLOCKD *const xd = &x->e_mbd;
+  struct optimize_ctx *const ctx = args->ctx;
+  struct macroblock_plane *const p = &x->plane[plane];
+  struct macroblockd_plane *const pd = &xd->plane[plane];
+  int16_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
+  int i, j;
+  uint8_t *dst;
+  MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
+  txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &i, &j);
+  dst = &pd->dst.buf[4 * j * pd->dst.stride + 4 * i];
+
+  // TODO(jingning): per transformed block zero forcing only enabled for
+  // luma component. will integrate chroma components as well.
+  if (x->zcoeff_blk[tx_size][block] && plane == 0) {
+    p->eobs[block] = 0;
+    ctx->ta[plane][i] = 0;
+    ctx->tl[plane][j] = 0;
+    return;
+  }
+
+  if (!x->skip_recode)
+    vp9_xform_quant(plane, block, plane_bsize, tx_size, arg);
+
+  if (x->optimize && (!x->skip_recode || !x->skip_optimize)) {
+    vp9_optimize_b(plane, block, plane_bsize, tx_size, x, ctx);
+  } else {
+    ctx->ta[plane][i] = p->eobs[block] > 0;
+    ctx->tl[plane][j] = p->eobs[block] > 0;
+  }
+
+  if (x->skip_encode || p->eobs[block] == 0)
+    return;
+
+  if (plane == PLANE_TYPE_Y) {
+    switch (mbmi->ext_txfrm) {
+      case NORM:
+        switch (tx_size) {
+          case TX_32X32:
+            vp9_idct32x32_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+            break;
+          case TX_16X16:
+            vp9_idct16x16_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+            break;
+          case TX_8X8:
+            vp9_idct8x8_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+            break;
+          case TX_4X4:
+            xd->itxm_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+            break;
+          default:
+            assert(0 && "Invalid transform size");
+        }
+        break;
+      case ALT:
+        switch (tx_size) {
+          case TX_32X32:
+#if CONFIG_EXT_TX_DST32
+            vp9_idst_add(dqcoeff, dst, pd->dst.stride, p->eobs[block], 32);
+#else
+            vp9_idct32x32_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+#endif
+            break;
+          case TX_16X16:
+            vp9_iht16x16_add(ADST_ADST, dqcoeff, dst, pd->dst.stride,
+                             p->eobs[block]);
+            break;
+          case TX_8X8:
+            vp9_iht8x8_add(ADST_ADST, dqcoeff, dst, pd->dst.stride,
+                           p->eobs[block]);
+            break;
+          case TX_4X4:
+            vp9_iht4x4_add(ADST_ADST, dqcoeff, dst, pd->dst.stride,
+                           p->eobs[block]);
+            break;
+          default:
+            assert(0 && "Invalid transform size");
+        }
+        break;
+      default:
+        assert(0 && "Invalid ext transform type");
+    }
+  } else {
+    switch (tx_size) {
+      case TX_32X32:
+        vp9_idct32x32_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+        break;
+      case TX_16X16:
+        vp9_idct16x16_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+        break;
+      case TX_8X8:
+        vp9_idct8x8_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+        break;
+      case TX_4X4:
+        // this is like vp9_short_idct4x4 but has a special case around eob<=1
+        // which is significant (not just an optimization) for the lossless
+        // case.
+        xd->itxm_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
+        break;
+      default:
+        assert(0 && "Invalid transform size");
+    }
+  }
+}
+#endif
 
 static void encode_block_pass1(int plane, int block, BLOCK_SIZE plane_bsize,
                                TX_SIZE tx_size, void *arg) {
@@ -485,6 +751,10 @@ static void encode_block_pass1(int plane, int block, BLOCK_SIZE plane_bsize,
   int16_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
   int i, j;
   uint8_t *dst;
+#if CONFIG_EXT_TX
+  MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
+  mbmi->ext_txfrm = NORM;
+#endif
   txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &i, &j);
   dst = &pd->dst.buf[4 * j * pd->dst.stride + 4 * i];
 
@@ -539,6 +809,9 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
   const scan_order *scan_order;
   TX_TYPE tx_type;
   MB_PREDICTION_MODE mode;
+#if CONFIG_FILTERINTRA
+  int fbit = 0;
+#endif
   const int bwl = b_width_log2(plane_bsize);
   const int diff_stride = 4 * (1 << bwl);
   uint8_t *src, *dst;
@@ -552,12 +825,21 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
 
   // if (x->optimize)
   // vp9_optimize_b(plane, block, plane_bsize, tx_size, x, args->ctx);
+#if CONFIG_FILTERINTRA
+      if (mbmi->sb_type < BLOCK_8X8 && plane == 0)
+        fbit = xd->mi_8x8[0]->b_filter_info[block];
+      else
+        fbit = plane == 0 ? mbmi->filterbit : mbmi->uv_filterbit;
+#endif
 
   switch (tx_size) {
     case TX_32X32:
       scan_order = &vp9_default_scan_orders[TX_32X32];
       mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
       vp9_predict_intra_block(xd, block >> 6, bwl, TX_32X32, mode,
+#if CONFIG_FILTERINTRA
+                              fbit,
+#endif
                               x->skip_encode ? src : dst,
                               x->skip_encode ? p->src.stride : pd->dst.stride,
                               dst, pd->dst.stride, i, j, plane);
@@ -581,6 +863,9 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
       scan_order = &vp9_scan_orders[TX_16X16][tx_type];
       mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
       vp9_predict_intra_block(xd, block >> 4, bwl, TX_16X16, mode,
+#if CONFIG_FILTERINTRA
+                              fbit,
+#endif
                               x->skip_encode ? src : dst,
                               x->skip_encode ? p->src.stride : pd->dst.stride,
                               dst, pd->dst.stride, i, j, plane);
@@ -601,6 +886,9 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
       scan_order = &vp9_scan_orders[TX_8X8][tx_type];
       mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
       vp9_predict_intra_block(xd, block >> 2, bwl, TX_8X8, mode,
+#if CONFIG_FILTERINTRA
+                              fbit,
+#endif
                               x->skip_encode ? src : dst,
                               x->skip_encode ? p->src.stride : pd->dst.stride,
                               dst, pd->dst.stride, i, j, plane);
@@ -625,6 +913,9 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
         mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
 
       vp9_predict_intra_block(xd, block, bwl, TX_4X4, mode,
+#if CONFIG_FILTERINTRA
+                              fbit,
+#endif
                               x->skip_encode ? src : dst,
                               x->skip_encode ? p->src.stride : pd->dst.stride,
                               dst, pd->dst.stride, i, j, plane);

vp9/encoder/vp9_firstpass.c

@@ -544,7 +544,6 @@ void vp9_first_pass(VP9_COMP *cpi) {
   }
   x->skip_recode = 0;
 
-
   // Initialise the MV cost table to the defaults
   // if( cm->current_video_frame == 0)
   // if ( 0 )

vp9/encoder/vp9_mbgraph.c

@@ -150,7 +150,13 @@ static int find_best_16x16_intra(VP9_COMP *cpi,
     unsigned int err;
 
     xd->mi_8x8[0]->mbmi.mode = mode;
+#if CONFIG_FILTERINTRA
+    xd->mi_8x8[0]->mbmi.filterbit = 0;
+#endif
     vp9_predict_intra_block(xd, 0, 2, TX_16X16, mode,
+#if CONFIG_FILTERINTRA
+                            0,
+#endif
                             x->plane[0].src.buf, x->plane[0].src.stride,
                             xd->plane[0].dst.buf, xd->plane[0].dst.stride,
                             0, 0, 0);

vp9/encoder/vp9_mcomp.c

@@ -2225,3 +2225,531 @@ int vp9_refining_search_8p_c(MACROBLOCK *x,
     return INT_MAX;
   }
 }
+
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+#define MVC(r, c)                                       \
+    (mvcost ?                                           \
+     ((mvjcost[((r) != rr) * 2 + ((c) != rc)] +         \
+       mvcost[0][((r) - rr)] + mvcost[1][((c) - rc)]) * \
+      error_per_bit + 4096) >> 13 : 0)
+
+#define SP(x) (((x) & 7) << 1)  // convert motion vector component to offset
+                                // for svf calc
+
+#define IFMVCV(r, c, s, e)                                \
+    if (c >= minc && c <= maxc && r >= minr && r <= maxr) \
+      s                                                   \
+    else                                                  \
+      e;
+
+/* pointer to predictor base of a motion vector */
+#define PRE(r, c) (y + (((r) >> 3) * y_stride + ((c) >> 3) -(offset)))
+
+/* returns subpixel variance error function */
+#define DIST(r, c) \
+    vfp->msvf(PRE(r, c), y_stride, SP(c), SP(r), z, src_stride, \
+              mask, mask_stride, &sse)
+
+/* checks if (r, c) has better score than previous best */
+#define CHECK_BETTER(v, r, c) \
+    IFMVCV(r, c, {                                                       \
+      thismse = (DIST(r, c));                                            \
+      if ((v = MVC(r, c) + thismse) < besterr) {                         \
+        besterr = v;                                                     \
+        br = r;                                                          \
+        bc = c;                                                          \
+        *distortion = thismse;                                           \
+        *sse1 = sse;                                                     \
+      }                                                                  \
+    },                                                                   \
+    v = INT_MAX;)
+
+int vp9_find_best_masked_sub_pixel_iterative(MACROBLOCK *x,
+                                      uint8_t *mask, int mask_stride,
+                                      MV *bestmv, const MV *ref_mv,
+                                      int allow_hp,
+                                      int error_per_bit,
+                                      const vp9_variance_fn_ptr_t *vfp,
+                                      int forced_stop,
+                                      int iters_per_step,
+                                      int *mvjcost, int *mvcost[2],
+                                      int *distortion,
+                                      unsigned int *sse1,
+                                      int is_second) {
+  uint8_t *z = x->plane[0].src.buf;
+  int src_stride = x->plane[0].src.stride;
+  MACROBLOCKD *xd = &x->e_mbd;
+
+  unsigned int besterr = INT_MAX;
+  unsigned int sse;
+  unsigned int whichdir;
+  unsigned int halfiters = iters_per_step;
+  unsigned int quarteriters = iters_per_step;
+  unsigned int eighthiters = iters_per_step;
+  int thismse;
+
+  const int y_stride = xd->plane[0].pre[is_second].stride;
+  const int offset = bestmv->row * y_stride + bestmv->col;
+  uint8_t *y = xd->plane[0].pre[is_second].buf + offset;
+
+  int rr = ref_mv->row;
+  int rc = ref_mv->col;
+  int br = bestmv->row * 8;
+  int bc = bestmv->col * 8;
+  int hstep = 4;
+  const int minc = MAX(x->mv_col_min * 8, ref_mv->col - MV_MAX);
+  const int maxc = MIN(x->mv_col_max * 8, ref_mv->col + MV_MAX);
+  const int minr = MAX(x->mv_row_min * 8, ref_mv->row - MV_MAX);
+  const int maxr = MIN(x->mv_row_max * 8, ref_mv->row + MV_MAX);
+
+  int tr = br;
+  int tc = bc;
+
+  // central mv
+  bestmv->row <<= 3;
+  bestmv->col <<= 3;
+
+  // calculate central point error
+  besterr = vfp->mvf(y, y_stride, z, src_stride, mask, mask_stride, sse1);
+  *distortion = besterr;
+  besterr += mv_err_cost(bestmv, ref_mv, mvjcost, mvcost, error_per_bit);
+
+  while (halfiters--) {
+    // 1/2 pel
+    FIRST_LEVEL_CHECKS;
+    // no reason to check the same one again.
+    if (tr == br && tc == bc)
+      break;
+    tr = br;
+    tc = bc;
+  }
+
+  // Note forced_stop: 0 - full, 1 - qtr only, 2 - half only
+  if (forced_stop != 2) {
+    hstep >>= 1;
+    while (quarteriters--) {
+      FIRST_LEVEL_CHECKS;
+      // no reason to check the same one again.
+      if (tr == br && tc == bc)
+        break;
+      tr = br;
+      tc = bc;
+    }
+  }
+
+  if (allow_hp && vp9_use_mv_hp(ref_mv) && forced_stop == 0) {
+    hstep >>= 1;
+    while (eighthiters--) {
+      FIRST_LEVEL_CHECKS;
+      // no reason to check the same one again.
+      if (tr == br && tc == bc)
+        break;
+      tr = br;
+      tc = bc;
+    }
+  }
+
+  bestmv->row = br;
+  bestmv->col = bc;
+
+  if ((abs(bestmv->col - ref_mv->col) > (MAX_FULL_PEL_VAL << 3)) ||
+      (abs(bestmv->row - ref_mv->row) > (MAX_FULL_PEL_VAL << 3)))
+    return INT_MAX;
+
+  return besterr;
+}
+
+int vp9_find_best_masked_sub_pixel_tree(MACROBLOCK *x,
+                                 uint8_t *mask, int mask_stride,
+                                 MV *bestmv, const MV *ref_mv,
+                                 int allow_hp,
+                                 int error_per_bit,
+                                 const vp9_variance_fn_ptr_t *vfp,
+                                 int forced_stop,
+                                 int iters_per_step,
+                                 int *mvjcost, int *mvcost[2],
+                                 int *distortion,
+                                 unsigned int *sse1,
+                                 int is_second) {
+  uint8_t *z = x->plane[0].src.buf;
+  const int src_stride = x->plane[0].src.stride;
+  MACROBLOCKD *xd = &x->e_mbd;
+  unsigned int besterr = INT_MAX;
+  unsigned int sse;
+  unsigned int whichdir;
+  int thismse;
+  unsigned int halfiters = iters_per_step;
+  unsigned int quarteriters = iters_per_step;
+  unsigned int eighthiters = iters_per_step;
+
+  const int y_stride = xd->plane[0].pre[is_second].stride;
+  const int offset = bestmv->row * y_stride + bestmv->col;
+  uint8_t *y = xd->plane[0].pre[is_second].buf + offset;
+
+  int rr = ref_mv->row;
+  int rc = ref_mv->col;
+  int br = bestmv->row * 8;
+  int bc = bestmv->col * 8;
+  int hstep = 4;
+  const int minc = MAX(x->mv_col_min * 8, ref_mv->col - MV_MAX);
+  const int maxc = MIN(x->mv_col_max * 8, ref_mv->col + MV_MAX);
+  const int minr = MAX(x->mv_row_min * 8, ref_mv->row - MV_MAX);
+  const int maxr = MIN(x->mv_row_max * 8, ref_mv->row + MV_MAX);
+
+  int tr = br;
+  int tc = bc;
+
+  // central mv
+  bestmv->row *= 8;
+  bestmv->col *= 8;
+
+  // calculate central point error
+  besterr = vfp->mvf(y, y_stride, z, src_stride, mask, mask_stride, sse1);
+  *distortion = besterr;
+  besterr += mv_err_cost(bestmv, ref_mv, mvjcost, mvcost, error_per_bit);
+
+  // 1/2 pel
+  FIRST_LEVEL_CHECKS;
+  if (halfiters > 1) {
+    SECOND_LEVEL_CHECKS;
+  }
+  tr = br;
+  tc = bc;
+
+  // Note forced_stop: 0 - full, 1 - qtr only, 2 - half only
+  if (forced_stop != 2) {
+    hstep >>= 1;
+    FIRST_LEVEL_CHECKS;
+    if (quarteriters > 1) {
+      SECOND_LEVEL_CHECKS;
+    }
+    tr = br;
+    tc = bc;
+  }
+
+  if (allow_hp && vp9_use_mv_hp(ref_mv) && forced_stop == 0) {
+    hstep >>= 1;
+    FIRST_LEVEL_CHECKS;
+    if (eighthiters > 1) {
+      SECOND_LEVEL_CHECKS;
+    }
+    tr = br;
+    tc = bc;
+  }
+
+  bestmv->row = br;
+  bestmv->col = bc;
+
+  if ((abs(bestmv->col - ref_mv->col) > (MAX_FULL_PEL_VAL << 3)) ||
+      (abs(bestmv->row - ref_mv->row) > (MAX_FULL_PEL_VAL << 3)))
+    return INT_MAX;
+
+  return besterr;
+}
+
+#undef MVC
+#undef SP
+#undef IFMVCV
+#undef PRE
+#undef DIST
+#undef CHECK_BETTER
+
+int vp9_masked_refining_search_sad_c(MACROBLOCK *x,
+                                     uint8_t * mask, int mask_stride,
+                                     MV *ref_mv, int error_per_bit,
+                                     int search_range,
+                                     vp9_variance_fn_ptr_t *fn_ptr,
+                                     int *mvjcost, int *mvcost[2],
+                                     const MV *center_mv, int is_second) {
+  const MACROBLOCKD* const xd = &x->e_mbd;
+  MV neighbors[4] = {{ -1, 0}, {0, -1}, {0, 1}, {1, 0}};
+  int i, j;
+  int this_row_offset, this_col_offset;
+
+  int what_stride = x->plane[0].src.stride;
+  int in_what_stride = xd->plane[0].pre[is_second].stride;
+  uint8_t *what = x->plane[0].src.buf;
+  uint8_t *best_address = xd->plane[0].pre[is_second].buf +
+                          (ref_mv->row * xd->plane[0].pre[is_second].stride) +
+                          ref_mv->col;
+  uint8_t *check_here;
+  unsigned int thissad;
+  MV this_mv;
+  unsigned int bestsad = INT_MAX;
+  MV fcenter_mv;
+
+  int *mvjsadcost = x->nmvjointsadcost;
+  int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]};
+
+  fcenter_mv.row = center_mv->row >> 3;
+  fcenter_mv.col = center_mv->col >> 3;
+
+  bestsad = fn_ptr->msdf(what, what_stride, best_address,
+                        in_what_stride, mask, mask_stride, 0x7fffffff) +
+                        mvsad_err_cost(ref_mv, &fcenter_mv,
+                                       mvjsadcost, mvsadcost, error_per_bit);
+
+  for (i = 0; i < search_range; i++) {
+    int best_site = -1;
+
+    for (j = 0; j < 4; j++) {
+      this_row_offset = ref_mv->row + neighbors[j].row;
+      this_col_offset = ref_mv->col + neighbors[j].col;
+
+      if ((this_col_offset > x->mv_col_min) &&
+          (this_col_offset < x->mv_col_max) &&
+          (this_row_offset > x->mv_row_min) &&
+          (this_row_offset < x->mv_row_max)) {
+        check_here = (neighbors[j].row) * in_what_stride + neighbors[j].col +
+                     best_address;
+        thissad = fn_ptr->msdf(what, what_stride, check_here, in_what_stride,
+                              mask, mask_stride, bestsad);
+
+        if (thissad < bestsad) {
+          this_mv.row = this_row_offset;
+          this_mv.col = this_col_offset;
+          thissad += mvsad_err_cost(&this_mv, &fcenter_mv,
+                                    mvjsadcost, mvsadcost, error_per_bit);
+
+          if (thissad < bestsad) {
+            bestsad = thissad;
+            best_site = j;
+          }
+        }
+      }
+    }
+
+    if (best_site == -1) {
+      break;
+    } else {
+      ref_mv->row += neighbors[best_site].row;
+      ref_mv->col += neighbors[best_site].col;
+      best_address += (neighbors[best_site].row) * in_what_stride +
+                      neighbors[best_site].col;
+    }
+  }
+
+  this_mv.row = ref_mv->row * 8;
+  this_mv.col = ref_mv->col * 8;
+
+  if (bestsad < INT_MAX)
+    return fn_ptr->mvf(what, what_stride, best_address, in_what_stride,
+                       mask, mask_stride,
+                      (unsigned int *)(&thissad)) +
+                      mv_err_cost(&this_mv, center_mv,
+                                  mvjcost, mvcost, x->errorperbit);
+  else
+    return INT_MAX;
+}
+
+int vp9_masked_diamond_search_sad_c(MACROBLOCK *x,
+                                    uint8_t *mask, int mask_stride,
+                                    MV *ref_mv, MV *best_mv,
+                                    int search_param, int sad_per_bit,
+                                    int *num00, vp9_variance_fn_ptr_t *fn_ptr,
+                                    int *mvjcost, int *mvcost[2],
+                                    const MV *center_mv, int is_second) {
+  int i, j, step;
+
+  const MACROBLOCKD* const xd = &x->e_mbd;
+  uint8_t *what = x->plane[0].src.buf;
+  int what_stride = x->plane[0].src.stride;
+  uint8_t *in_what;
+  int in_what_stride = xd->plane[0].pre[is_second].stride;
+  uint8_t *best_address;
+
+  int tot_steps;
+  MV this_mv;
+
+  int bestsad = INT_MAX;
+  int best_site = 0;
+  int last_site = 0;
+
+  int ref_row, ref_col;
+  int this_row_offset, this_col_offset;
+  search_site *ss;
+
+  uint8_t *check_here;
+  int thissad;
+  MV fcenter_mv;
+
+  int *mvjsadcost = x->nmvjointsadcost;
+  int *mvsadcost[2] = {x->nmvsadcost[0], x->nmvsadcost[1]};
+
+  fcenter_mv.row = center_mv->row >> 3;
+  fcenter_mv.col = center_mv->col >> 3;
+
+  clamp_mv(ref_mv, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
+  ref_row = ref_mv->row;
+  ref_col = ref_mv->col;
+  *num00 = 0;
+  best_mv->row = ref_row;
+  best_mv->col = ref_col;
+
+  // Work out the start point for the search
+  in_what = (uint8_t *)(xd->plane[0].pre[is_second].buf +
+                        ref_row * in_what_stride + ref_col);
+  best_address = in_what;
+
+  // Check the starting position
+  bestsad = fn_ptr->msdf(what, what_stride, in_what, in_what_stride,
+                         mask, mask_stride, 0x7fffffff)
+                + mvsad_err_cost(best_mv, &fcenter_mv,
+                                 mvjsadcost, mvsadcost, sad_per_bit);
+
+  // search_param determines the length of the initial step and hence the number
+  // of iterations
+  // 0 = initial step (MAX_FIRST_STEP) pel : 1 = (MAX_FIRST_STEP/2) pel, 2 =
+  // (MAX_FIRST_STEP/4) pel... etc.
+  ss = &x->ss[search_param * x->searches_per_step];
+  tot_steps = (x->ss_count / x->searches_per_step) - search_param;
+
+  i = 1;
+
+  for (step = 0; step < tot_steps; step++) {
+    for (j = 0; j < x->searches_per_step; j++) {
+      // Trap illegal vectors
+      this_row_offset = best_mv->row + ss[i].mv.row;
+      this_col_offset = best_mv->col + ss[i].mv.col;
+
+      if ((this_col_offset > x->mv_col_min) &&
+          (this_col_offset < x->mv_col_max) &&
+          (this_row_offset > x->mv_row_min) &&
+          (this_row_offset < x->mv_row_max)) {
+        check_here = ss[i].offset + best_address;
+        thissad = fn_ptr->msdf(what, what_stride, check_here, in_what_stride,
+                               mask, mask_stride, bestsad);
+
+        if (thissad < bestsad) {
+          this_mv.row = this_row_offset;
+          this_mv.col = this_col_offset;
+          thissad += mvsad_err_cost(&this_mv, &fcenter_mv,
+                                    mvjsadcost, mvsadcost, sad_per_bit);
+
+          if (thissad < bestsad) {
+            bestsad = thissad;
+            best_site = i;
+          }
+        }
+      }
+
+      i++;
+    }
+
+    if (best_site != last_site) {
+      best_mv->row += ss[best_site].mv.row;
+      best_mv->col += ss[best_site].mv.col;
+      best_address += ss[best_site].offset;
+      last_site = best_site;
+#if defined(NEW_DIAMOND_SEARCH)
+      while (1) {
+        this_row_offset = best_mv->row + ss[best_site].mv.row;
+        this_col_offset = best_mv->col + ss[best_site].mv.col;
+        if ((this_col_offset > x->mv_col_min) &&
+            (this_col_offset < x->mv_col_max) &&
+            (this_row_offset > x->mv_row_min) &&
+            (this_row_offset < x->mv_row_max)) {
+          check_here = ss[best_site].offset + best_address;
+          thissad = fn_ptr->msdf(what, what_stride, check_here, in_what_stride,
+                                 mask, mask_stride, bestsad);
+          if (thissad < bestsad) {
+            this_mv.row = this_row_offset;
+            this_mv.col = this_col_offset;
+            thissad += mvsad_err_cost(&this_mv, &fcenter_mv,
+                                      mvjsadcost, mvsadcost, sad_per_bit);
+            if (thissad < bestsad) {
+              bestsad = thissad;
+              best_mv->row += ss[best_site].mv.row;
+              best_mv->col += ss[best_site].mv.col;
+              best_address += ss[best_site].offset;
+              continue;
+            }
+          }
+        }
+        break;
+      };
+#endif
+    } else if (best_address == in_what) {
+      (*num00)++;
+    }
+  }
+
+  this_mv.row = best_mv->row * 8;
+  this_mv.col = best_mv->col * 8;
+
+  if (bestsad == INT_MAX)
+    return INT_MAX;
+
+  return fn_ptr->mvf(what, what_stride, best_address, in_what_stride,
+                     mask, mask_stride, (unsigned int *)(&thissad)) +
+                       mv_err_cost(&this_mv, center_mv,
+                                   mvjcost, mvcost, x->errorperbit);
+}
+
+int vp9_masked_full_pixel_diamond(VP9_COMP *cpi, MACROBLOCK *x,
+                                  uint8_t *mask, int mask_stride,
+                                  int_mv *mvp_full, int step_param,
+                                  int sadpb, int further_steps,
+                                  int do_refine, vp9_variance_fn_ptr_t *fn_ptr,
+                                  int_mv *ref_mv, int_mv *dst_mv,
+                                  int is_second) {
+  int_mv temp_mv;
+  int thissme, n, num00;
+  int bestsme = vp9_masked_diamond_search_sad_c(x, mask, mask_stride,
+                                              &mvp_full->as_mv, &temp_mv.as_mv,
+                                              step_param, sadpb, &num00,
+                                              fn_ptr, x->nmvjointcost,
+                                              x->mvcost, &ref_mv->as_mv,
+                                              is_second);
+  dst_mv->as_int = temp_mv.as_int;
+
+  n = num00;
+  num00 = 0;
+
+  /* If there won't be more n-step search, check to see if refining search is
+   * needed. */
+  if (n > further_steps)
+    do_refine = 0;
+
+  while (n < further_steps) {
+    n++;
+
+    if (num00) {
+      num00--;
+    } else {
+      thissme = vp9_masked_diamond_search_sad_c(x, mask, mask_stride,
+                                        &mvp_full->as_mv, &temp_mv.as_mv,
+                                        step_param + n, sadpb, &num00,
+                                        fn_ptr, x->nmvjointcost, x->mvcost,
+                                        &ref_mv->as_mv, is_second);
+
+      /* check to see if refining search is needed. */
+      if (num00 > (further_steps - n))
+        do_refine = 0;
+
+      if (thissme < bestsme) {
+        bestsme = thissme;
+        dst_mv->as_int = temp_mv.as_int;
+      }
+    }
+  }
+
+  /* final 1-away diamond refining search */
+  if (do_refine == 1) {
+    int search_range = 8;
+    int_mv best_mv;
+    best_mv.as_int = dst_mv->as_int;
+    thissme = vp9_masked_refining_search_sad_c(x, mask, mask_stride,
+                                       &best_mv.as_mv, sadpb, search_range,
+                                       fn_ptr, x->nmvjointcost, x->mvcost,
+                                       &ref_mv->as_mv, is_second);
+
+    if (thissme < bestsme) {
+      bestsme = thissme;
+      dst_mv->as_int = best_mv.as_int;
+    }
+  }
+  return bestsme;
+}
+#endif

vp9/encoder/vp9_mcomp.h

@@ -129,4 +129,38 @@ int vp9_refining_search_8p_c(MACROBLOCK *x,
                              int *mvjcost, int *mvcost[2],
                              const MV *center_mv, const uint8_t *second_pred,
                              int w, int h);
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+int vp9_find_best_masked_sub_pixel_iterative(MACROBLOCK *x,
+                                      uint8_t *mask, int mask_stride,
+                                      MV *bestmv, const MV *ref_mv,
+                                      int allow_hp,
+                                      int error_per_bit,
+                                      const vp9_variance_fn_ptr_t *vfp,
+                                      int forced_stop,
+                                      int iters_per_step,
+                                      int *mvjcost, int *mvcost[2],
+                                      int *distortion,
+                                      unsigned int *sse1,
+                                      int is_second);
+
+int vp9_find_best_masked_sub_pixel_tree(MACROBLOCK *x,
+                                      uint8_t *mask, int mask_stride,
+                                      MV *bestmv, const MV *ref_mv,
+                                      int allow_hp,
+                                      int error_per_bit,
+                                      const vp9_variance_fn_ptr_t *vfp,
+                                      int forced_stop,
+                                      int iters_per_step,
+                                      int *mvjcost, int *mvcost[2],
+                                      int *distortion,
+                                      unsigned int *sse1,
+                                      int is_second);
+
+int vp9_masked_full_pixel_diamond(
+    struct VP9_COMP *cpi, MACROBLOCK *x, uint8_t *mask, int mask_stride,
+    int_mv *mvp_full, int step_param, int sadpb, int further_steps,
+    int do_refine, vp9_variance_fn_ptr_t *fn_ptr, int_mv *ref_mv,
+    int_mv *dst_mv, int is_second);
+#endif
 #endif  // VP9_ENCODER_VP9_MCOMP_H_

vp9/encoder/vp9_onyx_if.c

@@ -521,24 +521,60 @@ static void set_rd_speed_thresholds(VP9_COMP *cpi, int mode) {
   sf->thresh_mult[THR_D207_PRED] += 2500;
   sf->thresh_mult[THR_D63_PRED] += 2500;
 
+#if CONFIG_INTERINTRA
+  sf->thresh_mult[THR_COMP_INTERINTRA_ZEROL   ] += 1500;
+  sf->thresh_mult[THR_COMP_INTERINTRA_ZEROG   ] += 1500;
+  sf->thresh_mult[THR_COMP_INTERINTRA_ZEROA   ] += 1500;
+
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEARESTL] += 1500;
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEARESTG] += 1500;
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEARESTA] += 1500;
+
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEARL   ] += 1500;
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEARG   ] += 1500;
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEARA   ] += 1500;
+
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEWL    ] += 2000;
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEWG    ] += 2000;
+  sf->thresh_mult[THR_COMP_INTERINTRA_NEWA    ] += 2000;
+#endif
+
   /* disable frame modes if flags not set */
   if (!(cpi->ref_frame_flags & VP9_LAST_FLAG)) {
     sf->thresh_mult[THR_NEWMV    ] = INT_MAX;
     sf->thresh_mult[THR_NEARESTMV] = INT_MAX;
     sf->thresh_mult[THR_ZEROMV   ] = INT_MAX;
     sf->thresh_mult[THR_NEARMV   ] = INT_MAX;
+#if CONFIG_INTERINTRA
+    sf->thresh_mult[THR_COMP_INTERINTRA_ZEROL   ] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEARESTL] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEARL   ] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEWL    ] = INT_MAX;
+#endif
   }
   if (!(cpi->ref_frame_flags & VP9_GOLD_FLAG)) {
     sf->thresh_mult[THR_NEARESTG ] = INT_MAX;
     sf->thresh_mult[THR_ZEROG    ] = INT_MAX;
     sf->thresh_mult[THR_NEARG    ] = INT_MAX;
     sf->thresh_mult[THR_NEWG     ] = INT_MAX;
+#if CONFIG_INTERINTRA
+    sf->thresh_mult[THR_COMP_INTERINTRA_ZEROG   ] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEARESTG] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEARG   ] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEWG    ] = INT_MAX;
+#endif
   }
   if (!(cpi->ref_frame_flags & VP9_ALT_FLAG)) {
     sf->thresh_mult[THR_NEARESTA ] = INT_MAX;
     sf->thresh_mult[THR_ZEROA    ] = INT_MAX;
     sf->thresh_mult[THR_NEARA    ] = INT_MAX;
     sf->thresh_mult[THR_NEWA     ] = INT_MAX;
+#if CONFIG_INTERINTRA
+    sf->thresh_mult[THR_COMP_INTERINTRA_ZEROA   ] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEARESTA] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEARA   ] = INT_MAX;
+    sf->thresh_mult[THR_COMP_INTERINTRA_NEWA    ] = INT_MAX;
+#endif
   }
 
   if ((cpi->ref_frame_flags & (VP9_LAST_FLAG | VP9_ALT_FLAG)) !=
@@ -1760,6 +1796,37 @@ VP9_PTR vp9_create_compressor(VP9_CONFIG *oxcf) {
       vp9_sub_pixel_avg_variance4x4, NULL, NULL, NULL,
       vp9_sad4x4x3, vp9_sad4x4x8, vp9_sad4x4x4d)
 
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+#define MBFP(BT, MSDF, MVF, MSVF) \
+  cpi->fn_ptr[BT].msdf            = MSDF; \
+  cpi->fn_ptr[BT].mvf             = MVF; \
+  cpi->fn_ptr[BT].msvf            = MSVF;
+
+  MBFP(BLOCK_64X64, vp9_masked_sad64x64, vp9_masked_variance64x64,
+       vp9_masked_sub_pixel_variance64x64)
+  MBFP(BLOCK_64X32, vp9_masked_sad64x32, vp9_masked_variance64x32,
+         vp9_masked_sub_pixel_variance64x32)
+  MBFP(BLOCK_32X64, vp9_masked_sad32x64, vp9_masked_variance32x64,
+         vp9_masked_sub_pixel_variance32x64)
+  MBFP(BLOCK_32X32, vp9_masked_sad32x32, vp9_masked_variance32x32,
+       vp9_masked_sub_pixel_variance32x32)
+  MBFP(BLOCK_32X16, vp9_masked_sad32x16, vp9_masked_variance32x16,
+       vp9_masked_sub_pixel_variance32x16)
+  MBFP(BLOCK_16X32, vp9_masked_sad16x32, vp9_masked_variance16x32,
+       vp9_masked_sub_pixel_variance16x32)
+  MBFP(BLOCK_16X16, vp9_masked_sad16x16, vp9_masked_variance16x16,
+         vp9_masked_sub_pixel_variance16x16)
+  MBFP(BLOCK_16X8, vp9_masked_sad16x8, vp9_masked_variance16x8,
+         vp9_masked_sub_pixel_variance16x8)
+  MBFP(BLOCK_8X16, vp9_masked_sad8x16, vp9_masked_variance8x16,
+         vp9_masked_sub_pixel_variance8x16)
+  MBFP(BLOCK_8X8, vp9_masked_sad8x8, vp9_masked_variance8x8,
+       vp9_masked_sub_pixel_variance8x8)
+  MBFP(BLOCK_4X4, vp9_masked_sad4x4, vp9_masked_variance4x4,
+         vp9_masked_sub_pixel_variance4x4)
+#endif
+
   cpi->full_search_sad = vp9_full_search_sad;
   cpi->diamond_search_sad = vp9_diamond_search_sad;
   cpi->refining_search_sad = vp9_refining_search_sad;
@@ -2234,6 +2301,44 @@ void vp9_write_yuv_rec_frame(VP9_COMMON *cm) {
 }
 #endif
 
+#if CONFIG_MASKED_INTERINTER
+static void select_masked_interinter_mode(VP9_COMP *cpi) {
+  static const double threshold = 1/128.0;
+  VP9_COMMON *cm = &cpi->common;
+  int sum = cpi->masked_interinter_select_counts[1] +
+      cpi->masked_interinter_select_counts[0];
+  if (sum) {
+    double fraction = (double) cpi->masked_interinter_select_counts[1] / sum;
+    cm->use_masked_interinter = (fraction > threshold);
+  }
+}
+#endif
+
+#if CONFIG_INTERINTRA
+static void select_interintra_mode(VP9_COMP *cpi) {
+  static const double threshold = 0.007;
+  VP9_COMMON *cm = &cpi->common;
+  int sum = cpi->interintra_select_count[1] + cpi->interintra_select_count[0];
+  if (sum) {
+    double fraction = (double)cpi->interintra_select_count[1] / (double)sum;
+    cm->use_interintra = (fraction > threshold);
+  }
+}
+
+#if CONFIG_MASKED_INTERINTRA
+static void select_masked_interintra_mode(VP9_COMP *cpi) {
+  static const double threshold = 1/100.0;
+  VP9_COMMON *cm = &cpi->common;
+  int sum = cpi->masked_interintra_select_count[1] +
+      cpi->masked_interintra_select_count[0];
+  if (sum) {
+    double fraction = (double) cpi->masked_interintra_select_count[1] / sum;
+    cm->use_masked_interintra = (fraction > threshold);
+  }
+}
+#endif
+#endif
+
 static void scale_and_extend_frame(YV12_BUFFER_CONFIG *src_fb,
                                    YV12_BUFFER_CONFIG *dst_fb) {
   const int in_w = src_fb->y_crop_width;
@@ -3023,6 +3128,21 @@ static void encode_frame_to_data_rate(VP9_COMP *cpi,
     set_high_precision_mv(cpi, (q < HIGH_PRECISION_MV_QTHRESH));
   }
 
+#if CONFIG_MASKED_INTERINTER
+  if (cm->current_video_frame == 0) {
+    cm->use_masked_interinter = 0;
+  }
+#endif
+
+#if CONFIG_INTERINTRA
+  if (cm->current_video_frame == 0) {
+    cm->use_interintra = 1;
+#if CONFIG_MASKED_INTERINTRA
+    cm->use_masked_interintra = 1;
+#endif
+  }
+#endif
+
   encode_with_recode_loop(cpi,
                           size,
                           dest,
@@ -3092,11 +3212,32 @@ static void encode_frame_to_data_rate(VP9_COMP *cpi,
     vp9_copy(counts->comp_inter, cpi->comp_inter_count);
     vp9_copy(counts->single_ref, cpi->single_ref_count);
     vp9_copy(counts->comp_ref, cpi->comp_ref_count);
+#if CONFIG_MASKED_INTERINTER
+    vp9_copy(counts->masked_interinter, cpi->masked_interinter_counts);
+#endif
+#if CONFIG_INTERINTRA
+    vp9_copy(counts->interintra, cpi->interintra_count);
+#if CONFIG_MASKED_INTERINTRA
+    vp9_copy(counts->masked_interintra, cpi->masked_interintra_count);
+#endif
+#endif
     if (!cpi->common.error_resilient_mode &&
         !cpi->common.frame_parallel_decoding_mode) {
       vp9_adapt_mode_probs(&cpi->common);
       vp9_adapt_mv_probs(&cpi->common, cpi->common.allow_high_precision_mv);
     }
+#if CONFIG_MASKED_INTERINTER
+    select_masked_interinter_mode(cpi);
+#endif
+#if CONFIG_INTERINTRA
+    select_interintra_mode(cpi);
+#if CONFIG_MASKED_INTERINTRA
+    if (cpi->common.use_interintra)
+      select_masked_interintra_mode(cpi);
+    else
+      cpi->common.use_masked_interintra = 0;
+#endif
+#endif
   }
 
 #ifdef ENTROPY_STATS

vp9/encoder/vp9_onyx_int.h

@@ -43,7 +43,11 @@
 
 #define KEY_FRAME_CONTEXT 5
 
+#if CONFIG_INTERINTRA
+#define MAX_MODES 42
+#else
 #define MAX_MODES 30
+#endif
 #define MAX_REFS  6
 
 #define MIN_THRESHMULT  32
@@ -161,6 +165,23 @@ typedef enum {
   THR_D63_PRED,
   THR_D117_PRED,
   THR_D45_PRED,
+
+#if CONFIG_INTERINTRA
+  THR_COMP_INTERINTRA_ZEROL,
+  THR_COMP_INTERINTRA_NEARESTL,
+  THR_COMP_INTERINTRA_NEARL,
+  THR_COMP_INTERINTRA_NEWL,
+
+  THR_COMP_INTERINTRA_ZEROG,
+  THR_COMP_INTERINTRA_NEARESTG,
+  THR_COMP_INTERINTRA_NEARG,
+  THR_COMP_INTERINTRA_NEWG,
+
+  THR_COMP_INTERINTRA_ZEROA,
+  THR_COMP_INTERINTRA_NEARESTA,
+  THR_COMP_INTERINTRA_NEARA,
+  THR_COMP_INTERINTRA_NEWA,
+#endif
 } THR_MODES;
 
 typedef enum {
@@ -471,6 +492,19 @@ typedef struct VP9_COMP {
   vp9_coeff_probs_model frame_coef_probs[TX_SIZES][PLANE_TYPES];
   vp9_coeff_stats frame_branch_ct[TX_SIZES][PLANE_TYPES];
 
+#if CONFIG_MASKED_INTERINTER
+  unsigned int masked_interinter_counts[BLOCK_SIZES][2];
+  unsigned int masked_interinter_select_counts[2];
+#endif
+#if CONFIG_INTERINTRA
+  unsigned int interintra_count[BLOCK_SIZES][2];
+  unsigned int interintra_select_count[2];
+#if CONFIG_MASKED_INTERINTRA
+  unsigned int masked_interintra_count[BLOCK_SIZES][2];
+  unsigned int masked_interintra_select_count[2];
+#endif
+#endif
+
   int64_t target_bandwidth;
   struct vpx_codec_pkt_list  *output_pkt_list;
 

vp9/encoder/vp9_sad_c.c

@@ -613,3 +613,127 @@ void vp9_sad4x4x4d_c(const uint8_t *src_ptr,
   sad_array[3] = vp9_sad4x4(src_ptr, src_stride,
                             ref_ptr[3], ref_stride, 0x7fffffff);
 }
+
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+unsigned int vp9_masked_sad64x64_c(const uint8_t *src_ptr,
+                                   int  src_stride,
+                                   const uint8_t *ref_ptr,
+                                   int  ref_stride,
+                                   const uint8_t *msk_ptr,
+                                   int  msk_stride,
+                                   unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride,  64, 64);
+}
+
+unsigned int vp9_masked_sad64x32_c(const uint8_t *src_ptr,
+                                   int  src_stride,
+                                   const uint8_t *ref_ptr,
+                                   int  ref_stride,
+                                   const uint8_t *msk_ptr,
+                                   int  msk_stride,
+                                   unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 64, 32);
+}
+
+unsigned int vp9_masked_sad32x64_c(const uint8_t *src_ptr,
+                                   int  src_stride,
+                                   const uint8_t *ref_ptr,
+                                   int  ref_stride,
+                                   const uint8_t *msk_ptr,
+                                   int  msk_stride,
+                                   unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 32, 64);
+}
+
+unsigned int vp9_masked_sad32x32_c(const uint8_t *src_ptr,
+                                   int  src_stride,
+                                   const uint8_t *ref_ptr,
+                                   int  ref_stride,
+                                   const uint8_t *msk_ptr,
+                                   int  msk_stride,
+                                   unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 32, 32);
+}
+
+unsigned int vp9_masked_sad16x32_c(const uint8_t *src_ptr,
+                                   int  src_stride,
+                                   const uint8_t *ref_ptr,
+                                   int  ref_stride,
+                                   const uint8_t *msk_ptr,
+                                   int  msk_stride,
+                                   unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 16, 32);
+}
+
+unsigned int vp9_masked_sad32x16_c(const uint8_t *src_ptr,
+                                   int  src_stride,
+                                   const uint8_t *ref_ptr,
+                                   int  ref_stride,
+                                   const uint8_t *msk_ptr,
+                                   int  msk_stride,
+                                   unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 32, 16);
+}
+
+unsigned int vp9_masked_sad16x16_c(const uint8_t *src_ptr,
+                                   int  src_stride,
+                                   const uint8_t *ref_ptr,
+                                   int  ref_stride,
+                                   const uint8_t *msk_ptr,
+                                   int  msk_stride,
+                                   unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 16, 16);
+}
+
+unsigned int vp9_masked_sad8x16_c(const uint8_t *src_ptr,
+                                  int  src_stride,
+                                  const uint8_t *ref_ptr,
+                                  int  ref_stride,
+                                  const uint8_t *msk_ptr,
+                                  int  msk_stride,
+                                  unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 8, 16);
+}
+
+unsigned int vp9_masked_sad16x8_c(const uint8_t *src_ptr,
+                                  int  src_stride,
+                                  const uint8_t *ref_ptr,
+                                  int  ref_stride,
+                                  const uint8_t *msk_ptr,
+                                  int  msk_stride,
+                                  unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 16, 8);
+}
+
+unsigned int vp9_masked_sad8x8_c(const uint8_t *src_ptr,
+                                 int  src_stride,
+                                 const uint8_t *ref_ptr,
+                                 int  ref_stride,
+                                 const uint8_t *msk_ptr,
+                                 int  msk_stride,
+                                 unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 8, 8);
+}
+
+unsigned int vp9_masked_sad4x4_c(const uint8_t *src_ptr,
+                                 int  src_stride,
+                                 const uint8_t *ref_ptr,
+                                 int  ref_stride,
+                                 const uint8_t *msk_ptr,
+                                 int  msk_stride,
+                                 unsigned int max_sad) {
+  return masked_sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride,
+                           msk_ptr, msk_stride, 4, 4);
+}
+#endif

vp9/encoder/vp9_sadmxn.h

@@ -35,4 +35,31 @@ static INLINE unsigned int sad_mx_n_c(const uint8_t *src_ptr,
   return sad;
 }
 
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+static INLINE unsigned int masked_sad_mx_n_c(const uint8_t *src_ptr,
+                                           int src_stride,
+                                           const uint8_t *ref_ptr,
+                                           int ref_stride,
+                                           const uint8_t *msk_ptr,
+                                           int msk_stride,
+                                           int m,
+                                           int n) {
+  int r, c;
+  unsigned int sad = 0;
+
+  for (r = 0; r < n; r++) {
+    for (c = 0; c < m; c++) {
+      sad += (msk_ptr[c]) * abs(src_ptr[c] - ref_ptr[c]);
+    }
+
+    src_ptr += src_stride;
+    ref_ptr += ref_stride;
+    msk_ptr += msk_stride;
+  }
+  sad = (sad + 31) >> 6;
+
+  return sad;
+}
+#endif
 #endif  // VP9_ENCODER_VP9_SADMXN_H_

vp9/encoder/vp9_variance.h

@@ -76,6 +76,33 @@ typedef unsigned int (*vp9_subp_avg_variance_fn_t)(const uint8_t *src_ptr,
                                                    unsigned int *sse,
                                                    const uint8_t *second_pred);
 
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+typedef unsigned int(*vp9_masked_sad_fn_t)(const uint8_t *src_ptr,
+                                         int source_stride,
+                                         const uint8_t *ref_ptr,
+                                         int ref_stride,
+                                         const uint8_t *msk_ptr,
+                                         int msk_stride,
+                                         unsigned int max_sad);
+typedef unsigned int (*vp9_masked_variance_fn_t)(const uint8_t *src_ptr,
+                                               int source_stride,
+                                               const uint8_t *ref_ptr,
+                                               int ref_stride,
+                                               const uint8_t *msk_ptr,
+                                               int msk_stride,
+                                               unsigned int *sse);
+typedef unsigned int (*vp9_masked_subpixvariance_fn_t)(const uint8_t *src_ptr,
+                                                     int source_stride,
+                                                     int xoffset,
+                                                     int yoffset,
+                                                     const uint8_t *ref_ptr,
+                                                     int Refstride,
+                                                     const uint8_t *msk_ptr,
+                                                     int msk_stride,
+                                                     unsigned int *sse);
+#endif
+
 typedef unsigned int (*vp9_getmbss_fn_t)(const short *);
 
 typedef unsigned int (*vp9_get16x16prederror_fn_t)(const uint8_t *src_ptr,
@@ -95,6 +122,12 @@ typedef struct vp9_variance_vtable {
   vp9_sad_multi_fn_t         sdx3f;
   vp9_sad_multi1_fn_t        sdx8f;
   vp9_sad_multi_d_fn_t       sdx4df;
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+  vp9_masked_sad_fn_t            msdf;
+  vp9_masked_variance_fn_t       mvf;
+  vp9_masked_subpixvariance_fn_t msvf;
+#endif
 } vp9_variance_fn_ptr_t;
 
 static void comp_avg_pred(uint8_t *comp_pred, const uint8_t *pred, int width,

vp9/encoder/vp9_variance_c.c

@@ -44,6 +44,40 @@ void variance(const uint8_t *src_ptr,
   }
 }
 
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+static void masked_variance(const uint8_t *src_ptr,
+                            int  src_stride,
+                            const uint8_t *ref_ptr,
+                            int  ref_stride,
+                            const uint8_t *msk_ptr,
+                            int  msk_stride,
+                            int  w,
+                            int  h,
+                            unsigned int *sse,
+                            int *sum) {
+  int i, j;
+  int diff;
+
+  *sum = 0;
+  *sse = 0;
+
+  for (i = 0; i < h; i++) {
+    for (j = 0; j < w; j++) {
+      diff = (src_ptr[j] - ref_ptr[j]) * (msk_ptr[j]);
+      *sum += diff;
+      *sse += diff * diff;
+    }
+
+    src_ptr += src_stride;
+    ref_ptr += ref_stride;
+    msk_ptr += msk_stride;
+  }
+  *sum = (*sum >= 0) ? ((*sum + 31) >> 6) : -((-*sum + 31) >> 6);
+  *sse = (*sse + 2047) >> 12;
+}
+#endif
+
 /****************************************************************************
  *
  *  ROUTINE       : filter_block2d_bil_first_pass
@@ -1092,3 +1126,449 @@ unsigned int vp9_sub_pixel_avg_variance4x8_c(const uint8_t *src_ptr,
   comp_avg_pred(temp3, second_pred, 4, 8, temp2, 4);
   return vp9_variance4x8(temp3, 4, dst_ptr, dst_pixels_per_line, sse);
 }
+
+#if ((CONFIG_MASKED_INTERINTRA && CONFIG_INTERINTRA) || \
+    CONFIG_MASKED_INTERINTER)
+unsigned int vp9_masked_variance64x64_c(const uint8_t *src_ptr,
+                                        int  source_stride,
+                                        const uint8_t *ref_ptr,
+                                        int  recon_stride,
+                                        const uint8_t *msk_ptr,
+                                        int  msk_stride,
+                                        unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 64, 64, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 12));
+}
+
+unsigned int vp9_masked_variance64x32_c(const uint8_t *src_ptr,
+                                        int  source_stride,
+                                        const uint8_t *ref_ptr,
+                                        int  recon_stride,
+                                        const uint8_t *msk_ptr,
+                                        int  msk_stride,
+                                        unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 64, 32, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 11));
+}
+
+unsigned int vp9_masked_variance32x64_c(const uint8_t *src_ptr,
+                                        int  source_stride,
+                                        const uint8_t *ref_ptr,
+                                        int  recon_stride,
+                                        const uint8_t *msk_ptr,
+                                        int  msk_stride,
+                                        unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 32, 64, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 11));
+}
+
+unsigned int vp9_masked_variance32x32_c(const uint8_t *src_ptr,
+                                        int  source_stride,
+                                        const uint8_t *ref_ptr,
+                                        int  recon_stride,
+                                        const uint8_t *msk_ptr,
+                                        int  msk_stride,
+                                        unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 32, 32, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 10));
+}
+
+unsigned int vp9_masked_variance32x16_c(const uint8_t *src_ptr,
+                                        int  source_stride,
+                                        const uint8_t *ref_ptr,
+                                        int  recon_stride,
+                                        const uint8_t *msk_ptr,
+                                        int  msk_stride,
+                                        unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 32, 16, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 9));
+}
+
+unsigned int vp9_masked_variance16x32_c(const uint8_t *src_ptr,
+                                        int  source_stride,
+                                        const uint8_t *ref_ptr,
+                                        int  recon_stride,
+                                        const uint8_t *msk_ptr,
+                                        int  msk_stride,
+                                        unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 16, 32, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 9));
+}
+
+unsigned int vp9_masked_variance16x16_c(const uint8_t *src_ptr,
+                                        int  source_stride,
+                                        const uint8_t *ref_ptr,
+                                        int  recon_stride,
+                                        const uint8_t *msk_ptr,
+                                        int  msk_stride,
+                                        unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 16, 16, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 8));
+}
+
+unsigned int vp9_masked_variance16x8_c(const uint8_t *src_ptr,
+                                       int  source_stride,
+                                       const uint8_t *ref_ptr,
+                                       int  recon_stride,
+                                       const uint8_t *msk_ptr,
+                                       int  msk_stride,
+                                       unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 16, 8, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 7));
+}
+
+unsigned int vp9_masked_variance8x16_c(const uint8_t *src_ptr,
+                                       int  source_stride,
+                                       const uint8_t *ref_ptr,
+                                       int  recon_stride,
+                                       const uint8_t *msk_ptr,
+                                       int  msk_stride,
+                                       unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 8, 16, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 7));
+}
+
+unsigned int vp9_masked_variance8x8_c(const uint8_t *src_ptr,
+                                      int  source_stride,
+                                      const uint8_t *ref_ptr,
+                                      int  recon_stride,
+                                      const uint8_t *msk_ptr,
+                                      int  msk_stride,
+                                      unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 8, 8, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 6));
+}
+
+unsigned int vp9_masked_variance4x4_c(const uint8_t *src_ptr,
+                                      int  source_stride,
+                                      const uint8_t *ref_ptr,
+                                      int  recon_stride,
+                                      const uint8_t *msk_ptr,
+                                      int  msk_stride,
+                                      unsigned int *sse) {
+  unsigned int var;
+  int avg;
+
+  masked_variance(src_ptr, source_stride, ref_ptr, recon_stride,
+                  msk_ptr, msk_stride, 4, 4, &var, &avg);
+  *sse = var;
+  return (var - (((int64_t)avg * avg) >> 4));
+}
+
+unsigned int vp9_masked_sub_pixel_variance64x64_c(const uint8_t *src_ptr,
+                                                  int  src_pixels_per_line,
+                                                  int  xoffset,
+                                                  int  yoffset,
+                                                  const uint8_t *dst_ptr,
+                                                  int dst_pixels_per_line,
+                                                  const uint8_t *msk_ptr,
+                                                  int msk_stride,
+                                                  unsigned int *sse) {
+  uint16_t fdata3[65 * 64];  // Temp data bufffer used in filtering
+  uint8_t temp2[68 * 64];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 65, 64, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 64, 64, 64, 64, vfilter);
+
+  return vp9_masked_variance64x64_c(temp2, 64, dst_ptr, dst_pixels_per_line,
+                                    msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance64x32_c(const uint8_t *src_ptr,
+                                                  int  src_pixels_per_line,
+                                                  int  xoffset,
+                                                  int  yoffset,
+                                                  const uint8_t *dst_ptr,
+                                                  int dst_pixels_per_line,
+                                                  const uint8_t *msk_ptr,
+                                                  int msk_stride,
+                                                  unsigned int *sse) {
+  uint16_t fdata3[65 * 64];  // Temp data bufffer used in filtering
+  uint8_t temp2[68 * 64];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 33, 64, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 64, 64, 32, 64, vfilter);
+
+  return vp9_masked_variance64x32_c(temp2, 64, dst_ptr, dst_pixels_per_line,
+                                    msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance32x64_c(const uint8_t *src_ptr,
+                                                  int  src_pixels_per_line,
+                                                  int  xoffset,
+                                                  int  yoffset,
+                                                  const uint8_t *dst_ptr,
+                                                  int dst_pixels_per_line,
+                                                  const uint8_t *msk_ptr,
+                                                  int msk_stride,
+                                                  unsigned int *sse) {
+  uint16_t fdata3[65 * 64];  // Temp data bufffer used in filtering
+  uint8_t temp2[68 * 64];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 65, 32, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 64, 32, vfilter);
+
+  return vp9_masked_variance32x64_c(temp2, 32, dst_ptr, dst_pixels_per_line,
+                                    msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance32x32_c(const uint8_t *src_ptr,
+                                                  int  src_pixels_per_line,
+                                                  int  xoffset,
+                                                  int  yoffset,
+                                                  const uint8_t *dst_ptr,
+                                                  int dst_pixels_per_line,
+                                                  const uint8_t *msk_ptr,
+                                                  int msk_stride,
+                                                  unsigned int *sse) {
+  uint16_t fdata3[33 * 32];  // Temp data bufffer used in filtering
+  uint8_t temp2[36 * 32];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 33, 32, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 32, 32, vfilter);
+
+  return vp9_masked_variance32x32_c(temp2, 32, dst_ptr, dst_pixels_per_line,
+                                    msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance32x16_c(const uint8_t *src_ptr,
+                                           int  src_pixels_per_line,
+                                           int  xoffset,
+                                           int  yoffset,
+                                           const uint8_t *dst_ptr,
+                                           int dst_pixels_per_line,
+                                           const uint8_t *msk_ptr,
+                                           int msk_stride,
+                                           unsigned int *sse) {
+  uint16_t fdata3[33 * 32];  // Temp data bufffer used in filtering
+  uint8_t temp2[36 * 32];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 17, 32, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 16, 32, vfilter);
+
+  return vp9_masked_variance32x16_c(temp2, 32, dst_ptr, dst_pixels_per_line,
+                                    msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance16x32_c(const uint8_t *src_ptr,
+                                                  int  src_pixels_per_line,
+                                                  int  xoffset,
+                                                  int  yoffset,
+                                                  const uint8_t *dst_ptr,
+                                                  int dst_pixels_per_line,
+                                                  const uint8_t *msk_ptr,
+                                                  int msk_stride,
+                                                  unsigned int *sse) {
+  uint16_t fdata3[33 * 32];  // Temp data bufffer used in filtering
+  uint8_t temp2[36 * 32];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 33, 16, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 32, 16, vfilter);
+
+  return vp9_masked_variance16x32_c(temp2, 16, dst_ptr, dst_pixels_per_line,
+                                    msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance16x16_c(const uint8_t *src_ptr,
+                                                  int  src_pixels_per_line,
+                                                  int  xoffset,
+                                                  int  yoffset,
+                                                  const uint8_t *dst_ptr,
+                                                  int dst_pixels_per_line,
+                                                  const uint8_t *msk_ptr,
+                                                  int msk_stride,
+                                                  unsigned int *sse) {
+  uint16_t fdata3[17 * 16];  // Temp data bufffer used in filtering
+  uint8_t temp2[20 * 16];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 17, 16, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 16, 16, vfilter);
+
+  return vp9_masked_variance16x16_c(temp2, 16, dst_ptr, dst_pixels_per_line,
+                                    msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance16x8_c(const uint8_t *src_ptr,
+                                                 int  src_pixels_per_line,
+                                                 int  xoffset,
+                                                 int  yoffset,
+                                                 const uint8_t *dst_ptr,
+                                                 int dst_pixels_per_line,
+                                                 const uint8_t *msk_ptr,
+                                                 int msk_stride,
+                                                 unsigned int *sse) {
+  uint16_t fdata3[16 * 9];  // Temp data bufffer used in filtering
+  uint8_t temp2[20 * 16];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 9, 16, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 8, 16, vfilter);
+
+  return vp9_masked_variance16x8_c(temp2, 16, dst_ptr, dst_pixels_per_line,
+                                   msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance8x16_c(const uint8_t *src_ptr,
+                                                 int  src_pixels_per_line,
+                                                 int  xoffset,
+                                                 int  yoffset,
+                                                 const uint8_t *dst_ptr,
+                                                 int dst_pixels_per_line,
+                                                 const uint8_t *msk_ptr,
+                                                 int msk_stride,
+                                                 unsigned int *sse) {
+  uint16_t fdata3[9 * 16];  // Temp data bufffer used in filtering
+  uint8_t temp2[20 * 16];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 17, 8, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 16, 8, vfilter);
+
+  return vp9_masked_variance8x16_c(temp2, 8, dst_ptr, dst_pixels_per_line,
+                                   msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance8x8_c(const uint8_t *src_ptr,
+                                                int  src_pixels_per_line,
+                                                int  xoffset,
+                                                int  yoffset,
+                                                const uint8_t *dst_ptr,
+                                                int dst_pixels_per_line,
+                                                const uint8_t *msk_ptr,
+                                                int msk_stride,
+                                                unsigned int *sse) {
+  uint16_t fdata3[9 * 8];  // Temp data bufffer used in filtering
+  uint8_t temp2[20 * 16];
+  const int16_t *hfilter, *vfilter;
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 9, 8, hfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 8, 8, 8, 8, vfilter);
+
+  return vp9_masked_variance8x8_c(temp2, 8, dst_ptr, dst_pixels_per_line,
+                                  msk_ptr, msk_stride, sse);
+}
+
+unsigned int vp9_masked_sub_pixel_variance4x4_c(const uint8_t *src_ptr,
+                                                int  src_pixels_per_line,
+                                                int  xoffset,
+                                                int  yoffset,
+                                                const uint8_t *dst_ptr,
+                                                int dst_pixels_per_line,
+                                                const uint8_t *msk_ptr,
+                                                int msk_stride,
+                                                unsigned int *sse) {
+  uint8_t temp2[20 * 16];
+  const int16_t *hfilter, *vfilter;
+  uint16_t fdata3[5 * 4];  // Temp data bufffer used in filtering
+
+  hfilter = BILINEAR_FILTERS_2TAP(xoffset);
+  vfilter = BILINEAR_FILTERS_2TAP(yoffset);
+
+  // First filter 1d Horizontal
+  var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
+                                    1, 5, 4, hfilter);
+
+  // Now filter Verticaly
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 4,  4,  4,  4, vfilter);
+
+  return vp9_masked_variance4x4_c(temp2, 4, dst_ptr, dst_pixels_per_line,
+                                  msk_ptr, msk_stride, sse);
+}
+#endif

vpxenc.c

@@ -44,6 +44,10 @@
 #include "./webmenc.h"
 #include "./y4minput.h"
 
+#if CONFIG_EXT_TX
+extern int block_index;
+#endif
+
 /* Swallow warnings about unused results of fread/fwrite */
 static size_t wrap_fread(void *ptr, size_t size, size_t nmemb,
                          FILE *stream) {