Merge "Fixes to buffer update for temporal layers."
diff --git a/configure b/configure
index 621161c..729c986 100755
--- a/configure
+++ b/configure
@@ -249,7 +249,6 @@
     unistd_h
 "
 EXPERIMENT_LIST="
-    oneshotq
     multiple_arf
     non420
     alpha
diff --git a/vp9/common/vp9_blockd.h b/vp9/common/vp9_blockd.h
index ff047b8..d0d4852 100644
--- a/vp9/common/vp9_blockd.h
+++ b/vp9/common/vp9_blockd.h
@@ -417,7 +417,7 @@
   *y = (raster_mb >> tx_cols_log2) << tx_size;
 }
 
-static void extend_for_intra(MACROBLOCKD* const xd, BLOCK_SIZE plane_bsize,
+static void extend_for_intra(MACROBLOCKD *xd, BLOCK_SIZE plane_bsize,
                              int plane, int block, TX_SIZE tx_size) {
   struct macroblockd_plane *const pd = &xd->plane[plane];
   uint8_t *const buf = pd->dst.buf;
@@ -457,7 +457,7 @@
     }
   }
 }
-static void set_contexts_on_border(MACROBLOCKD *xd,
+static void set_contexts_on_border(const MACROBLOCKD *xd,
                                    struct macroblockd_plane *pd,
                                    BLOCK_SIZE plane_bsize,
                                    int tx_size_in_blocks, int has_eob,
@@ -495,7 +495,7 @@
     L[pt] = 0;
 }
 
-static void set_contexts(MACROBLOCKD *xd, struct macroblockd_plane *pd,
+static void set_contexts(const MACROBLOCKD *xd, struct macroblockd_plane *pd,
                          BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
                          int has_eob, int aoff, int loff) {
   ENTROPY_CONTEXT *const A = pd->above_context + aoff;
diff --git a/vp9/common/vp9_entropy.c b/vp9/common/vp9_entropy.c
index 2640ac7..d3a867c 100644
--- a/vp9/common/vp9_entropy.c
+++ b/vp9/common/vp9_entropy.c
@@ -322,9 +322,8 @@
   vp9_coeff_count_model *coef_counts = cm->counts.coef[tx_size];
   unsigned int (*eob_branch_count)[REF_TYPES][COEF_BANDS][PREV_COEF_CONTEXTS] =
       cm->counts.eob_branch[tx_size];
-  int t, i, j, k, l;
+  int i, j, k, l, m;
   unsigned int branch_ct[UNCONSTRAINED_NODES][2];
-  vp9_prob coef_probs[UNCONSTRAINED_NODES];
 
   for (i = 0; i < BLOCK_TYPES; ++i)
     for (j = 0; j < REF_TYPES; ++j)
@@ -332,15 +331,14 @@
         for (l = 0; l < PREV_COEF_CONTEXTS; ++l) {
           if (l >= 3 && k == 0)
             continue;
-          vp9_tree_probs_from_distribution(vp9_coefmodel_tree, coef_probs,
-                                           branch_ct, coef_counts[i][j][k][l],
-                                           0);
+          vp9_tree_probs_from_distribution(vp9_coefmodel_tree, branch_ct,
+                                           coef_counts[i][j][k][l], 0);
           branch_ct[0][1] = eob_branch_count[i][j][k][l] - branch_ct[0][0];
-          coef_probs[0] = get_binary_prob(branch_ct[0][0], branch_ct[0][1]);
-          for (t = 0; t < UNCONSTRAINED_NODES; ++t)
-            dst_coef_probs[i][j][k][l][t] = merge_probs(
-                pre_coef_probs[i][j][k][l][t], coef_probs[t],
-                branch_ct[t], count_sat, update_factor);
+          for (m = 0; m < UNCONSTRAINED_NODES; ++m)
+            dst_coef_probs[i][j][k][l][m] = merge_probs(
+                                                pre_coef_probs[i][j][k][l][m],
+                                                branch_ct[m],
+                                                count_sat, update_factor);
         }
 }
 
diff --git a/vp9/common/vp9_entropy.h b/vp9/common/vp9_entropy.h
index ec7d09a..c58e852 100644
--- a/vp9/common/vp9_entropy.h
+++ b/vp9/common/vp9_entropy.h
@@ -153,8 +153,8 @@
 
 void vp9_model_to_full_probs(const vp9_prob *model, vp9_prob *full);
 
-static int get_entropy_context(TX_SIZE tx_size,
-                               ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l) {
+static int get_entropy_context(TX_SIZE tx_size, const ENTROPY_CONTEXT *a,
+                                                const ENTROPY_CONTEXT *l) {
   ENTROPY_CONTEXT above_ec = 0, left_ec = 0;
 
   switch (tx_size) {
@@ -163,16 +163,16 @@
       left_ec = l[0] != 0;
       break;
     case TX_8X8:
-      above_ec = !!*(uint16_t *)a;
-      left_ec  = !!*(uint16_t *)l;
+      above_ec = !!*(const uint16_t *)a;
+      left_ec  = !!*(const uint16_t *)l;
       break;
     case TX_16X16:
-      above_ec = !!*(uint32_t *)a;
-      left_ec  = !!*(uint32_t *)l;
+      above_ec = !!*(const uint32_t *)a;
+      left_ec  = !!*(const uint32_t *)l;
       break;
     case TX_32X32:
-      above_ec = !!*(uint64_t *)a;
-      left_ec  = !!*(uint64_t *)l;
+      above_ec = !!*(const uint64_t *)a;
+      left_ec  = !!*(const uint64_t *)l;
       break;
     default:
       assert(!"Invalid transform size.");
diff --git a/vp9/common/vp9_entropymode.c b/vp9/common/vp9_entropymode.c
index 21c91d6..a963d55 100644
--- a/vp9/common/vp9_entropymode.c
+++ b/vp9/common/vp9_entropymode.c
@@ -161,51 +161,52 @@
   { 101,  21, 107, 181, 192, 103,  19,  67, 125 }   // y = tm
 };
 
-static const vp9_prob default_partition_probs[FRAME_TYPES][PARTITION_CONTEXTS]
+const vp9_prob vp9_kf_partition_probs[PARTITION_CONTEXTS]
+                                     [PARTITION_TYPES - 1] = {
+  // 8x8 -> 4x4
+  { 158,  97,  94 },  // a/l both not split
+  {  93,  24,  99 },  // a split, l not split
+  {  85, 119,  44 },  // l split, a not split
+  {  62,  59,  67 },  // a/l both split
+  // 16x16 -> 8x8
+  { 149,  53,  53 },  // a/l both not split
+  {  94,  20,  48 },  // a split, l not split
+  {  83,  53,  24 },  // l split, a not split
+  {  52,  18,  18 },  // a/l both split
+  // 32x32 -> 16x16
+  { 150,  40,  39 },  // a/l both not split
+  {  78,  12,  26 },  // a split, l not split
+  {  67,  33,  11 },  // l split, a not split
+  {  24,   7,   5 },  // a/l both split
+  // 64x64 -> 32x32
+  { 174,  35,  49 },  // a/l both not split
+  {  68,  11,  27 },  // a split, l not split
+  {  57,  15,   9 },  // l split, a not split
+  {  12,   3,   3 },  // a/l both split
+};
+
+static const vp9_prob default_partition_probs[PARTITION_CONTEXTS]
                                              [PARTITION_TYPES - 1] = {
-  {  // frame_type = keyframe
-    // 8x8 -> 4x4
-    { 158,  97,  94 },  // a/l both not split
-    {  93,  24,  99 },  // a split, l not split
-    {  85, 119,  44 },  // l split, a not split
-    {  62,  59,  67 },  // a/l both split
-    // 16x16 -> 8x8
-    { 149,  53,  53 },  // a/l both not split
-    {  94,  20,  48 },  // a split, l not split
-    {  83,  53,  24 },  // l split, a not split
-    {  52,  18,  18 },  // a/l both split
-    // 32x32 -> 16x16
-    { 150,  40,  39 },  // a/l both not split
-    {  78,  12,  26 },  // a split, l not split
-    {  67,  33,  11 },  // l split, a not split
-    {  24,   7,   5 },  // a/l both split
-    // 64x64 -> 32x32
-    { 174,  35,  49 },  // a/l both not split
-    {  68,  11,  27 },  // a split, l not split
-    {  57,  15,   9 },  // l split, a not split
-    {  12,   3,   3 },  // a/l both split
-  }, {  // frame_type = interframe
-    // 8x8 -> 4x4
-    { 199, 122, 141 },  // a/l both not split
-    { 147,  63, 159 },  // a split, l not split
-    { 148, 133, 118 },  // l split, a not split
-    { 121, 104, 114 },  // a/l both split
-    // 16x16 -> 8x8
-    { 174,  73,  87 },  // a/l both not split
-    {  92,  41,  83 },  // a split, l not split
-    {  82,  99,  50 },  // l split, a not split
-    {  53,  39,  39 },  // a/l both split
-    // 32x32 -> 16x16
-    { 177,  58,  59 },  // a/l both not split
-    {  68,  26,  63 },  // a split, l not split
-    {  52,  79,  25 },  // l split, a not split
-    {  17,  14,  12 },  // a/l both split
-    // 64x64 -> 32x32
-    { 222,  34,  30 },  // a/l both not split
-    {  72,  16,  44 },  // a split, l not split
-    {  58,  32,  12 },  // l split, a not split
-    {  10,   7,   6 },  // a/l both split
-  }
+  // 8x8 -> 4x4
+  { 199, 122, 141 },  // a/l both not split
+  { 147,  63, 159 },  // a split, l not split
+  { 148, 133, 118 },  // l split, a not split
+  { 121, 104, 114 },  // a/l both split
+  // 16x16 -> 8x8
+  { 174,  73,  87 },  // a/l both not split
+  {  92,  41,  83 },  // a split, l not split
+  {  82,  99,  50 },  // l split, a not split
+  {  53,  39,  39 },  // a/l both split
+  // 32x32 -> 16x16
+  { 177,  58,  59 },  // a/l both not split
+  {  68,  26,  63 },  // a split, l not split
+  {  52,  79,  25 },  // l split, a not split
+  {  17,  14,  12 },  // a/l both split
+  // 64x64 -> 32x32
+  { 222,  34,  30 },  // a/l both not split
+  {  72,  16,  44 },  // a split, l not split
+  {  58,  32,  12 },  // l split, a not split
+  {  10,   7,   6 },  // a/l both split
 };
 
 static const vp9_prob default_inter_mode_probs[INTER_MODE_CONTEXTS]
@@ -349,29 +350,15 @@
 #define COUNT_SAT 20
 #define MAX_UPDATE_FACTOR 128
 
-static int update_ct(vp9_prob pre_prob, vp9_prob prob,
-                     const unsigned int ct[2]) {
-  return merge_probs(pre_prob, prob, ct, COUNT_SAT, MAX_UPDATE_FACTOR);
+static int adapt_prob(vp9_prob pre_prob, const unsigned int ct[2]) {
+  return merge_probs(pre_prob, ct, COUNT_SAT, MAX_UPDATE_FACTOR);
 }
 
-static int update_ct2(vp9_prob pre_prob, const unsigned int ct[2]) {
-  return merge_probs2(pre_prob, ct, COUNT_SAT, MAX_UPDATE_FACTOR);
-}
-
-static void update_mode_probs(int n_modes,
-                              const vp9_tree_index *tree,
-                              const unsigned int *cnt,
-                              const vp9_prob *pre_probs, vp9_prob *dst_probs,
-                              unsigned int tok0_offset) {
-#define MAX_PROBS 32
-  vp9_prob probs[MAX_PROBS];
-  unsigned int branch_ct[MAX_PROBS][2];
-  int t;
-
-  assert(n_modes - 1 < MAX_PROBS);
-  vp9_tree_probs_from_distribution(tree, probs, branch_ct, cnt, tok0_offset);
-  for (t = 0; t < n_modes - 1; ++t)
-    dst_probs[t] = update_ct(pre_probs[t], probs[t], branch_ct[t]);
+static void adapt_probs(const vp9_tree_index *tree,
+                        const vp9_prob *pre_probs, const unsigned int *counts,
+                        unsigned int offset, vp9_prob *probs) {
+  tree_merge_probs(tree, pre_probs, counts, offset,
+                   COUNT_SAT, MAX_UPDATE_FACTOR, probs);
 }
 
 void vp9_adapt_mode_probs(VP9_COMMON *cm) {
@@ -381,46 +368,40 @@
   const FRAME_COUNTS *counts = &cm->counts;
 
   for (i = 0; i < INTRA_INTER_CONTEXTS; i++)
-    fc->intra_inter_prob[i] = update_ct2(pre_fc->intra_inter_prob[i],
+    fc->intra_inter_prob[i] = adapt_prob(pre_fc->intra_inter_prob[i],
                                          counts->intra_inter[i]);
   for (i = 0; i < COMP_INTER_CONTEXTS; i++)
-    fc->comp_inter_prob[i] = update_ct2(pre_fc->comp_inter_prob[i],
+    fc->comp_inter_prob[i] = adapt_prob(pre_fc->comp_inter_prob[i],
                                         counts->comp_inter[i]);
   for (i = 0; i < REF_CONTEXTS; i++)
-    fc->comp_ref_prob[i] = update_ct2(pre_fc->comp_ref_prob[i],
+    fc->comp_ref_prob[i] = adapt_prob(pre_fc->comp_ref_prob[i],
                                       counts->comp_ref[i]);
   for (i = 0; i < REF_CONTEXTS; i++)
     for (j = 0; j < 2; j++)
-      fc->single_ref_prob[i][j] = update_ct2(pre_fc->single_ref_prob[i][j],
+      fc->single_ref_prob[i][j] = adapt_prob(pre_fc->single_ref_prob[i][j],
                                              counts->single_ref[i][j]);
 
   for (i = 0; i < INTER_MODE_CONTEXTS; i++)
-    update_mode_probs(INTER_MODES, vp9_inter_mode_tree,
-                      counts->inter_mode[i], pre_fc->inter_mode_probs[i],
-                      fc->inter_mode_probs[i], NEARESTMV);
+    adapt_probs(vp9_inter_mode_tree, pre_fc->inter_mode_probs[i],
+                counts->inter_mode[i], NEARESTMV, fc->inter_mode_probs[i]);
 
   for (i = 0; i < BLOCK_SIZE_GROUPS; i++)
-    update_mode_probs(INTRA_MODES, vp9_intra_mode_tree,
-                      counts->y_mode[i], pre_fc->y_mode_prob[i],
-                      fc->y_mode_prob[i], 0);
+    adapt_probs(vp9_intra_mode_tree, pre_fc->y_mode_prob[i],
+                counts->y_mode[i], 0, fc->y_mode_prob[i]);
 
   for (i = 0; i < INTRA_MODES; ++i)
-    update_mode_probs(INTRA_MODES, vp9_intra_mode_tree,
-                      counts->uv_mode[i], pre_fc->uv_mode_prob[i],
-                      fc->uv_mode_prob[i], 0);
+    adapt_probs(vp9_intra_mode_tree, pre_fc->uv_mode_prob[i],
+                counts->uv_mode[i], 0, fc->uv_mode_prob[i]);
 
   for (i = 0; i < PARTITION_CONTEXTS; i++)
-    update_mode_probs(PARTITION_TYPES, vp9_partition_tree,
-                      counts->partition[i],
-                      pre_fc->partition_prob[INTER_FRAME][i],
-                      fc->partition_prob[INTER_FRAME][i], 0);
+    adapt_probs(vp9_partition_tree, pre_fc->partition_prob[i],
+                counts->partition[i], 0, fc->partition_prob[i]);
 
   if (cm->mcomp_filter_type == SWITCHABLE) {
     for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++)
-      update_mode_probs(SWITCHABLE_FILTERS, vp9_switchable_interp_tree,
-                        counts->switchable_interp[i],
-                        pre_fc->switchable_interp_prob[i],
-                        fc->switchable_interp_prob[i], 0);
+      adapt_probs(vp9_switchable_interp_tree, pre_fc->switchable_interp_prob[i],
+                  counts->switchable_interp[i], 0,
+                  fc->switchable_interp_prob[i]);
   }
 
   if (cm->tx_mode == TX_MODE_SELECT) {
@@ -432,23 +413,23 @@
     for (i = 0; i < TX_SIZE_CONTEXTS; ++i) {
       tx_counts_to_branch_counts_8x8(counts->tx.p8x8[i], branch_ct_8x8p);
       for (j = 0; j < TX_SIZES - 3; ++j)
-        fc->tx_probs.p8x8[i][j] = update_ct2(pre_fc->tx_probs.p8x8[i][j],
+        fc->tx_probs.p8x8[i][j] = adapt_prob(pre_fc->tx_probs.p8x8[i][j],
                                              branch_ct_8x8p[j]);
 
       tx_counts_to_branch_counts_16x16(counts->tx.p16x16[i], branch_ct_16x16p);
       for (j = 0; j < TX_SIZES - 2; ++j)
-        fc->tx_probs.p16x16[i][j] = update_ct2(pre_fc->tx_probs.p16x16[i][j],
+        fc->tx_probs.p16x16[i][j] = adapt_prob(pre_fc->tx_probs.p16x16[i][j],
                                                branch_ct_16x16p[j]);
 
       tx_counts_to_branch_counts_32x32(counts->tx.p32x32[i], branch_ct_32x32p);
       for (j = 0; j < TX_SIZES - 1; ++j)
-        fc->tx_probs.p32x32[i][j] = update_ct2(pre_fc->tx_probs.p32x32[i][j],
+        fc->tx_probs.p32x32[i][j] = adapt_prob(pre_fc->tx_probs.p32x32[i][j],
                                                branch_ct_32x32p[j]);
     }
   }
 
   for (i = 0; i < MBSKIP_CONTEXTS; ++i)
-    fc->mbskip_probs[i] = update_ct2(pre_fc->mbskip_probs[i],
+    fc->mbskip_probs[i] = adapt_prob(pre_fc->mbskip_probs[i],
                                      counts->mbskip[i]);
 }
 
diff --git a/vp9/common/vp9_entropymode.h b/vp9/common/vp9_entropymode.h
index ea96555..38b4199 100644
--- a/vp9/common/vp9_entropymode.h
+++ b/vp9/common/vp9_entropymode.h
@@ -38,6 +38,9 @@
 extern const vp9_prob vp9_kf_y_mode_prob[INTRA_MODES][INTRA_MODES]
                                         [INTRA_MODES - 1];
 
+extern const vp9_prob vp9_kf_partition_probs[PARTITION_CONTEXTS]
+                                            [PARTITION_TYPES - 1];
+
 extern const vp9_tree_index vp9_intra_mode_tree[TREE_SIZE(INTRA_MODES)];
 extern struct vp9_token vp9_intra_mode_encodings[INTRA_MODES];
 
diff --git a/vp9/common/vp9_entropymv.c b/vp9/common/vp9_entropymv.c
index f70b571..b061cdb 100644
--- a/vp9/common/vp9_entropymv.c
+++ b/vp9/common/vp9_entropymv.c
@@ -191,60 +191,47 @@
 }
 
 static vp9_prob adapt_prob(vp9_prob prep, const unsigned int ct[2]) {
-  return merge_probs2(prep, ct, MV_COUNT_SAT, MV_MAX_UPDATE_FACTOR);
+  return merge_probs(prep, ct, MV_COUNT_SAT, MV_MAX_UPDATE_FACTOR);
 }
 
-static unsigned int adapt_probs(unsigned int i,
-                                vp9_tree tree,
-                                vp9_prob this_probs[],
-                                const vp9_prob last_probs[],
-                                const unsigned int num_events[]) {
-  const unsigned int left = tree[i] <= 0
-          ? num_events[-tree[i]]
-          : adapt_probs(tree[i], tree, this_probs, last_probs, num_events);
-
-  const unsigned int right = tree[i + 1] <= 0
-          ? num_events[-tree[i + 1]]
-          : adapt_probs(tree[i + 1], tree, this_probs, last_probs, num_events);
-  const unsigned int ct[2] = { left, right };
-  this_probs[i >> 1] = adapt_prob(last_probs[i >> 1], ct);
-  return left + right;
+static void adapt_probs(const vp9_tree_index *tree, const vp9_prob *pre_probs,
+                        const unsigned int *counts, vp9_prob *probs) {
+  tree_merge_probs(tree, pre_probs, counts, 0,
+                   MV_COUNT_SAT, MV_MAX_UPDATE_FACTOR, probs);
 }
 
-
 void vp9_adapt_mv_probs(VP9_COMMON *cm, int allow_hp) {
   int i, j;
 
-  const FRAME_CONTEXT *pre_fc = &cm->frame_contexts[cm->frame_context_idx];
+  nmv_context *fc = &cm->fc.nmvc;
+  const nmv_context *pre_fc = &cm->frame_contexts[cm->frame_context_idx].nmvc;
+  const nmv_context_counts *counts = &cm->counts.mv;
 
-  nmv_context *ctx = &cm->fc.nmvc;
-  const nmv_context *pre_ctx = &pre_fc->nmvc;
-  const nmv_context_counts *cts = &cm->counts.mv;
-
-  adapt_probs(0, vp9_mv_joint_tree, ctx->joints, pre_ctx->joints, cts->joints);
+  adapt_probs(vp9_mv_joint_tree, pre_fc->joints, counts->joints,
+                     fc->joints);
 
   for (i = 0; i < 2; ++i) {
-    ctx->comps[i].sign = adapt_prob(pre_ctx->comps[i].sign, cts->comps[i].sign);
-    adapt_probs(0, vp9_mv_class_tree, ctx->comps[i].classes,
-                pre_ctx->comps[i].classes, cts->comps[i].classes);
-    adapt_probs(0, vp9_mv_class0_tree, ctx->comps[i].class0,
-                pre_ctx->comps[i].class0, cts->comps[i].class0);
+    nmv_component *comp = &fc->comps[i];
+    const nmv_component *pre_comp = &pre_fc->comps[i];
+    const nmv_component_counts *c = &counts->comps[i];
+
+    comp->sign = adapt_prob(pre_comp->sign, c->sign);
+    adapt_probs(vp9_mv_class_tree, pre_comp->classes, c->classes,
+                comp->classes);
+    adapt_probs(vp9_mv_class0_tree, pre_comp->class0, c->class0, comp->class0);
 
     for (j = 0; j < MV_OFFSET_BITS; ++j)
-        ctx->comps[i].bits[j] = adapt_prob(pre_ctx->comps[i].bits[j],
-                                           cts->comps[i].bits[j]);
+      comp->bits[j] = adapt_prob(pre_comp->bits[j], c->bits[j]);
 
     for (j = 0; j < CLASS0_SIZE; ++j)
-      adapt_probs(0, vp9_mv_fp_tree, ctx->comps[i].class0_fp[j],
-                  pre_ctx->comps[i].class0_fp[j], cts->comps[i].class0_fp[j]);
+      adapt_probs(vp9_mv_fp_tree, pre_comp->class0_fp[j], c->class0_fp[j],
+                  comp->class0_fp[j]);
 
-    adapt_probs(0, vp9_mv_fp_tree, ctx->comps[i].fp, pre_ctx->comps[i].fp,
-                cts->comps[i].fp);
+    adapt_probs(vp9_mv_fp_tree, pre_comp->fp, c->fp, comp->fp);
 
     if (allow_hp) {
-      ctx->comps[i].class0_hp = adapt_prob(pre_ctx->comps[i].class0_hp,
-                                           cts->comps[i].class0_hp);
-      ctx->comps[i].hp = adapt_prob(pre_ctx->comps[i].hp, cts->comps[i].hp);
+      comp->class0_hp = adapt_prob(pre_comp->class0_hp, c->class0_hp);
+      comp->hp = adapt_prob(pre_comp->hp, c->hp);
     }
   }
 }
diff --git a/vp9/common/vp9_filter.c b/vp9/common/vp9_filter.c
index 8f24052..79ace14 100644
--- a/vp9/common/vp9_filter.c
+++ b/vp9/common/vp9_filter.c
@@ -97,19 +97,15 @@
   { 0, -3,  1,  38, 64, 32, -1, -3}
 };
 
+
+static const subpel_kernel* vp9_filter_kernels[4] = {
+  vp9_sub_pel_filters_8,
+  vp9_sub_pel_filters_8lp,
+  vp9_sub_pel_filters_8s,
+  vp9_bilinear_filters
+};
+
 const subpel_kernel *vp9_get_filter_kernel(INTERPOLATION_TYPE type) {
-  switch (type) {
-    case EIGHTTAP:
-      return vp9_sub_pel_filters_8;
-    case EIGHTTAP_SMOOTH:
-      return vp9_sub_pel_filters_8lp;
-    case EIGHTTAP_SHARP:
-      return vp9_sub_pel_filters_8s;
-    case BILINEAR:
-      return vp9_bilinear_filters;
-    default:
-      assert(!"Invalid interpolation type.");
-      return NULL;
-  }
+  return vp9_filter_kernels[type];
 }
 
diff --git a/vp9/common/vp9_filter.h b/vp9/common/vp9_filter.h
index 8652a6e..b1e7e64 100644
--- a/vp9/common/vp9_filter.h
+++ b/vp9/common/vp9_filter.h
@@ -39,7 +39,6 @@
 const subpel_kernel *vp9_get_filter_kernel(INTERPOLATION_TYPE type);
 
 extern const subpel_kernel vp9_bilinear_filters[SUBPEL_SHIFTS];
-extern const subpel_kernel vp9_sub_pel_filters_6[SUBPEL_SHIFTS];
 extern const subpel_kernel vp9_sub_pel_filters_8[SUBPEL_SHIFTS];
 extern const subpel_kernel vp9_sub_pel_filters_8s[SUBPEL_SHIFTS];
 extern const subpel_kernel vp9_sub_pel_filters_8lp[SUBPEL_SHIFTS];
diff --git a/vp9/common/vp9_onyxc_int.h b/vp9/common/vp9_onyxc_int.h
index ba2e9d8..a2af57a 100644
--- a/vp9/common/vp9_onyxc_int.h
+++ b/vp9/common/vp9_onyxc_int.h
@@ -41,7 +41,7 @@
 typedef struct frame_contexts {
   vp9_prob y_mode_prob[BLOCK_SIZE_GROUPS][INTRA_MODES - 1];
   vp9_prob uv_mode_prob[INTRA_MODES][INTRA_MODES - 1];
-  vp9_prob partition_prob[FRAME_TYPES][PARTITION_CONTEXTS][PARTITION_TYPES - 1];
+  vp9_prob partition_prob[PARTITION_CONTEXTS][PARTITION_TYPES - 1];
   vp9_coeff_probs_model coef_probs[TX_SIZES][BLOCK_TYPES];
   vp9_prob switchable_interp_prob[SWITCHABLE_FILTER_CONTEXTS]
                                  [SWITCHABLE_FILTERS - 1];
@@ -245,6 +245,11 @@
   return ALIGN_POWER_OF_TWO(n_mis, MI_BLOCK_SIZE_LOG2);
 }
 
+static INLINE const vp9_prob* get_partition_probs(VP9_COMMON *cm, int ctx) {
+  return cm->frame_type == KEY_FRAME ? vp9_kf_partition_probs[ctx]
+                                     : cm->fc.partition_prob[ctx];
+}
+
 static INLINE void set_skip_context(
     MACROBLOCKD *xd,
     ENTROPY_CONTEXT *above_context[MAX_MB_PLANE],
diff --git a/vp9/common/vp9_reconintra.c b/vp9/common/vp9_reconintra.c
index bd609dc..eb643b0 100644
--- a/vp9/common/vp9_reconintra.c
+++ b/vp9/common/vp9_reconintra.c
@@ -369,7 +369,7 @@
   }
 }
 
-void vp9_predict_intra_block(MACROBLOCKD *xd, int block_idx, int bwl_in,
+void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
                             TX_SIZE tx_size, int mode,
                             const uint8_t *ref, int ref_stride,
                             uint8_t *dst, int dst_stride) {
diff --git a/vp9/common/vp9_reconintra.h b/vp9/common/vp9_reconintra.h
index e9d0dbf..6e3f55c 100644
--- a/vp9/common/vp9_reconintra.h
+++ b/vp9/common/vp9_reconintra.h
@@ -14,8 +14,8 @@
 #include "vpx/vpx_integer.h"
 #include "vp9/common/vp9_blockd.h"
 
-void vp9_predict_intra_block(MACROBLOCKD *xd, int block_idx, int bwl_in,
-                            TX_SIZE tx_size, int mode,
-                            const uint8_t *ref, int ref_stride,
-                            uint8_t *dst, int dst_stride);
+void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
+                             TX_SIZE tx_size, int mode,
+                             const uint8_t *ref, int ref_stride,
+                             uint8_t *dst, int dst_stride);
 #endif  // VP9_COMMON_VP9_RECONINTRA_H_
diff --git a/vp9/common/vp9_scan.h b/vp9/common/vp9_scan.h
index a5c8463..14a1a7e 100644
--- a/vp9/common/vp9_scan.h
+++ b/vp9/common/vp9_scan.h
@@ -191,8 +191,7 @@
 }
 
 static INLINE int get_coef_context(const int16_t *neighbors,
-                                   uint8_t *token_cache,
-                                   int c) {
+                                   const uint8_t *token_cache, int c) {
   return (1 + token_cache[neighbors[MAX_NEIGHBORS * c + 0]] +
           token_cache[neighbors[MAX_NEIGHBORS * c + 1]]) >> 1;
 }
diff --git a/vp9/common/vp9_treecoder.c b/vp9/common/vp9_treecoder.c
index da1213d..1805fb4 100644
--- a/vp9/common/vp9_treecoder.c
+++ b/vp9/common/vp9_treecoder.c
@@ -40,9 +40,7 @@
   tree2tok(p - offset, t, 0, 0, 0);
 }
 
-static unsigned int convert_distribution(unsigned int i,
-                                         vp9_tree tree,
-                                         vp9_prob probs[],
+static unsigned int convert_distribution(unsigned int i, vp9_tree tree,
                                          unsigned int branch_ct[][2],
                                          const unsigned int num_events[],
                                          unsigned int tok0_offset) {
@@ -51,24 +49,25 @@
   if (tree[i] <= 0) {
     left = num_events[-tree[i] - tok0_offset];
   } else {
-    left = convert_distribution(tree[i], tree, probs, branch_ct,
-                                num_events, tok0_offset);
+    left = convert_distribution(tree[i], tree, branch_ct, num_events,
+                                tok0_offset);
   }
   if (tree[i + 1] <= 0)
     right = num_events[-tree[i + 1] - tok0_offset];
   else
-    right = convert_distribution(tree[i + 1], tree, probs, branch_ct,
-                                 num_events, tok0_offset);
+    right = convert_distribution(tree[i + 1], tree, branch_ct, num_events,
+                                 tok0_offset);
 
-  probs[i>>1] = get_binary_prob(left, right);
-  branch_ct[i>>1][0] = left;
-  branch_ct[i>>1][1] = right;
+  branch_ct[i >> 1][0] = left;
+  branch_ct[i >> 1][1] = right;
   return left + right;
 }
 
-void vp9_tree_probs_from_distribution(vp9_tree tree, vp9_prob probs[/* n-1 */],
+void vp9_tree_probs_from_distribution(vp9_tree tree,
                                       unsigned int branch_ct[/* n-1 */][2],
                                       const unsigned int num_events[/* n */],
                                       unsigned int tok0_offset) {
-  convert_distribution(0, tree, probs, branch_ct, num_events, tok0_offset);
+  convert_distribution(0, tree, branch_ct, num_events, tok0_offset);
 }
+
+
diff --git a/vp9/common/vp9_treecoder.h b/vp9/common/vp9_treecoder.h
index 4ba171f..9c776d6 100644
--- a/vp9/common/vp9_treecoder.h
+++ b/vp9/common/vp9_treecoder.h
@@ -50,11 +50,11 @@
    probability updates. */
 
 void vp9_tree_probs_from_distribution(vp9_tree tree,
-                                      vp9_prob probs[ /* n - 1 */ ],
                                       unsigned int branch_ct[ /* n - 1 */ ][2],
                                       const unsigned int num_events[ /* n */ ],
                                       unsigned int tok0_offset);
 
+
 static INLINE vp9_prob clip_prob(int p) {
   return (p > 255) ? 255u : (p < 1) ? 1u : p;
 }
@@ -81,21 +81,46 @@
   return ROUND_POWER_OF_TWO(prob1 * (256 - factor) + prob2 * factor, 8);
 }
 
-static INLINE vp9_prob merge_probs(vp9_prob pre_prob, vp9_prob prob,
+static INLINE vp9_prob merge_probs(vp9_prob pre_prob,
                                    const unsigned int ct[2],
                                    unsigned int count_sat,
                                    unsigned int max_update_factor) {
+  const vp9_prob prob = get_binary_prob(ct[0], ct[1]);
   const unsigned int count = MIN(ct[0] + ct[1], count_sat);
   const unsigned int factor = max_update_factor * count / count_sat;
   return weighted_prob(pre_prob, prob, factor);
 }
 
-static INLINE vp9_prob merge_probs2(vp9_prob pre_prob,
-                                   const unsigned int ct[2],
-                                   unsigned int count_sat,
-                                   unsigned int max_update_factor) {
-  return merge_probs(pre_prob, get_binary_prob(ct[0], ct[1]), ct, count_sat,
-                     max_update_factor);
+static unsigned int tree_merge_probs_impl(unsigned int i,
+                                          const vp9_tree_index *tree,
+                                          const vp9_prob *pre_probs,
+                                          const unsigned int *counts,
+                                          unsigned int count_sat,
+                                          unsigned int max_update_factor,
+                                          vp9_prob *probs) {
+  const int l = tree[i];
+  const unsigned int left_count = (l <= 0)
+                 ? counts[-l]
+                 : tree_merge_probs_impl(l, tree, pre_probs, counts,
+                                         count_sat, max_update_factor, probs);
+  const int r = tree[i + 1];
+  const unsigned int right_count = (r <= 0)
+                 ? counts[-r]
+                 : tree_merge_probs_impl(r, tree, pre_probs, counts,
+                                         count_sat, max_update_factor, probs);
+  const unsigned int ct[2] = { left_count, right_count };
+  probs[i >> 1] = merge_probs(pre_probs[i >> 1], ct,
+                              count_sat, max_update_factor);
+  return left_count + right_count;
+}
+
+static void tree_merge_probs(const vp9_tree_index *tree,
+                             const vp9_prob *pre_probs,
+                             const unsigned int *counts, int offset,
+                             unsigned int count_sat,
+                             unsigned int max_update_factor, vp9_prob *probs) {
+  tree_merge_probs_impl(0, tree, pre_probs, &counts[-offset],
+                        count_sat, max_update_factor, probs);
 }
 
 
diff --git a/vp9/common/x86/vp9_intrapred_ssse3.asm b/vp9/common/x86/vp9_intrapred_ssse3.asm
index 568e208..88df9b2 100644
--- a/vp9/common/x86/vp9_intrapred_ssse3.asm
+++ b/vp9/common/x86/vp9_intrapred_ssse3.asm
@@ -991,7 +991,7 @@
   lea              dst8q, [dst8q+strideq*4]
 
   ; output 2nd half of 3rd 8 lines and half of 4th 8 lines
-  mova                m0, [sh_b23456789abcdefff]
+  mova                m0, [GLOBAL(sh_b23456789abcdefff)]
   mova  [dstq           +16], m7
   mova  [dst8q             ], m7
   pshufb              m7, m0
diff --git a/vp9/decoder/vp9_decodemv.c b/vp9/decoder/vp9_decodemv.c
index 475a299..1ca5786 100644
--- a/vp9/decoder/vp9_decodemv.c
+++ b/vp9/decoder/vp9_decodemv.c
@@ -149,16 +149,17 @@
   return segment_id;
 }
 
-static uint8_t read_skip_coeff(VP9_COMMON *const cm, MACROBLOCKD *const xd,
-                               int segment_id, vp9_reader *r) {
-  int skip_coeff = vp9_segfeature_active(&cm->seg, segment_id, SEG_LVL_SKIP);
-  if (!skip_coeff) {
+static int read_skip_coeff(VP9_COMMON *cm, const MACROBLOCKD *xd,
+                           int segment_id, vp9_reader *r) {
+  if (vp9_segfeature_active(&cm->seg, segment_id, SEG_LVL_SKIP)) {
+    return 1;
+  } else {
     const int ctx = vp9_get_pred_context_mbskip(xd);
-    skip_coeff = vp9_read(r, vp9_get_pred_prob_mbskip(cm, xd));
+    const int skip = vp9_read(r, cm->fc.mbskip_probs[ctx]);
     if (!cm->frame_parallel_decoding_mode)
-      ++cm->counts.mbskip[ctx][skip_coeff];
+      ++cm->counts.mbskip[ctx][skip];
+    return skip;
   }
-  return skip_coeff;
 }
 
 static void read_intra_frame_mode_info(VP9_COMMON *const cm,
diff --git a/vp9/decoder/vp9_decodframe.c b/vp9/decoder/vp9_decodframe.c
index 41d1400..1fd9e97 100644
--- a/vp9/decoder/vp9_decodframe.c
+++ b/vp9/decoder/vp9_decodframe.c
@@ -41,6 +41,7 @@
   VP9_COMMON *cm;
   vp9_reader bit_reader;
   DECLARE_ALIGNED(16, MACROBLOCKD, xd);
+  DECLARE_ALIGNED(16, unsigned char, token_cache[1024]);
 } TileWorkerData;
 
 static int read_be32(const uint8_t *p) {
@@ -297,6 +298,7 @@
   VP9_COMMON *cm;
   MACROBLOCKD *xd;
   vp9_reader *r;
+  unsigned char* token_cache;
 };
 
 static void predict_and_reconstruct_intra_block(int plane, int block,
@@ -326,7 +328,7 @@
 
   if (!mi->mbmi.skip_coeff) {
     vp9_decode_block_tokens(cm, xd, plane, block, plane_bsize, tx_size,
-                            args->r);
+                            args->r, args->token_cache);
     inverse_transform_block(xd, plane, block, plane_bsize, tx_size);
   }
 }
@@ -336,6 +338,7 @@
   MACROBLOCKD *xd;
   vp9_reader *r;
   int *eobtotal;
+  unsigned char* token_cache;
 };
 
 static void reconstruct_inter_block(int plane, int block,
@@ -346,7 +349,8 @@
   MACROBLOCKD *const xd = args->xd;
 
   *args->eobtotal += vp9_decode_block_tokens(cm, xd, plane, block,
-                                             plane_bsize, tx_size, args->r);
+                                             plane_bsize, tx_size,
+                                             args->r, args->token_cache);
   inverse_transform_block(xd, plane, block, plane_bsize, tx_size);
 }
 
@@ -398,7 +402,8 @@
 static void decode_modes_b(VP9_COMMON *const cm, MACROBLOCKD *const xd,
                            const TileInfo *const tile,
                            int mi_row, int mi_col,
-                           vp9_reader *r, BLOCK_SIZE bsize) {
+                           vp9_reader *r, BLOCK_SIZE bsize,
+                           unsigned char *token_cache) {
   const int less8x8 = bsize < BLOCK_8X8;
   MB_MODE_INFO *mbmi;
 
@@ -420,7 +425,7 @@
   }
 
   if (!is_inter_block(mbmi)) {
-    struct intra_args arg = { cm, xd, r };
+    struct intra_args arg = { cm, xd, r, token_cache };
     foreach_transformed_block(xd, bsize, predict_and_reconstruct_intra_block,
                               &arg);
   } else {
@@ -438,7 +443,7 @@
     // Reconstruction
     if (!mbmi->skip_coeff) {
       int eobtotal = 0;
-      struct inter_args arg = { cm, xd, r, &eobtotal };
+      struct inter_args arg = { cm, xd, r, &eobtotal, token_cache };
       foreach_transformed_block(xd, bsize, reconstruct_inter_block, &arg);
       if (!less8x8 && eobtotal == 0)
         mbmi->skip_coeff = 1;  // skip loopfilter
@@ -448,66 +453,74 @@
   xd->corrupted |= vp9_reader_has_error(r);
 }
 
-static PARTITION_TYPE read_partition(int hbs, int mi_rows, int mi_cols,
-                                     int mi_row, int mi_col,
-                                     vp9_prob probs[PARTITION_TYPES - 1],
+static PARTITION_TYPE read_partition(VP9_COMMON *cm, MACROBLOCKD *xd, int hbs,
+                                     int mi_row, int mi_col, BLOCK_SIZE bsize,
                                      vp9_reader *r) {
-  const int has_rows = (mi_row + hbs) < mi_rows;
-  const int has_cols = (mi_col + hbs) < mi_cols;
+  const int ctx = partition_plane_context(xd->above_seg_context,
+                                          xd->left_seg_context,
+                                          mi_row, mi_col, bsize);
+  const vp9_prob *const probs = get_partition_probs(cm, ctx);
+  const int has_rows = (mi_row + hbs) < cm->mi_rows;
+  const int has_cols = (mi_col + hbs) < cm->mi_cols;
+  PARTITION_TYPE p;
 
   if (has_rows && has_cols)
-    return treed_read(r, vp9_partition_tree, probs);
+    p = treed_read(r, vp9_partition_tree, probs);
   else if (!has_rows && has_cols)
-    return vp9_read(r, probs[1]) ? PARTITION_SPLIT : PARTITION_HORZ;
+    p = vp9_read(r, probs[1]) ? PARTITION_SPLIT : PARTITION_HORZ;
   else if (has_rows && !has_cols)
-    return vp9_read(r, probs[2]) ? PARTITION_SPLIT : PARTITION_VERT;
+    p = vp9_read(r, probs[2]) ? PARTITION_SPLIT : PARTITION_VERT;
   else
-    return PARTITION_SPLIT;
+    p = PARTITION_SPLIT;
+
+  if (!cm->frame_parallel_decoding_mode)
+    ++cm->counts.partition[ctx][p];
+
+  return p;
 }
 
 static void decode_modes_sb(VP9_COMMON *const cm, MACROBLOCKD *const xd,
                             const TileInfo *const tile,
                             int mi_row, int mi_col,
-                            vp9_reader* r, BLOCK_SIZE bsize) {
+                            vp9_reader* r, BLOCK_SIZE bsize,
+                            unsigned char *token_cache) {
   const int hbs = num_8x8_blocks_wide_lookup[bsize] / 2;
   PARTITION_TYPE partition;
   BLOCK_SIZE subsize;
-  int ctx;
 
   if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
     return;
 
-  ctx = partition_plane_context(xd->above_seg_context, xd->left_seg_context,
-                                mi_row, mi_col, bsize);
-  partition = read_partition(hbs, cm->mi_rows, cm->mi_cols, mi_row, mi_col,
-                             cm->fc.partition_prob[cm->frame_type][ctx], r);
-
-  if (!cm->frame_parallel_decoding_mode)
-    ++cm->counts.partition[ctx][partition];
-
+  partition = read_partition(cm, xd, hbs, mi_row, mi_col, bsize, r);
   subsize = get_subsize(bsize, partition);
   if (subsize < BLOCK_8X8) {
-    decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize);
+    decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize, token_cache);
   } else {
     switch (partition) {
       case PARTITION_NONE:
-        decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize);
+        decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize, token_cache);
         break;
       case PARTITION_HORZ:
-        decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize);
+        decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize, token_cache);
         if (mi_row + hbs < cm->mi_rows)
-          decode_modes_b(cm, xd, tile, mi_row + hbs, mi_col, r, subsize);
+          decode_modes_b(cm, xd, tile, mi_row + hbs, mi_col, r, subsize,
+                         token_cache);
         break;
       case PARTITION_VERT:
-        decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize);
+        decode_modes_b(cm, xd, tile, mi_row, mi_col, r, subsize, token_cache);
         if (mi_col + hbs < cm->mi_cols)
-          decode_modes_b(cm, xd, tile, mi_row, mi_col + hbs, r, subsize);
+          decode_modes_b(cm, xd, tile, mi_row, mi_col + hbs, r, subsize,
+                         token_cache);
         break;
       case PARTITION_SPLIT:
-        decode_modes_sb(cm, xd, tile, mi_row, mi_col, r, subsize);
-        decode_modes_sb(cm, xd, tile, mi_row, mi_col + hbs, r, subsize);
-        decode_modes_sb(cm, xd, tile, mi_row + hbs, mi_col, r, subsize);
-        decode_modes_sb(cm, xd, tile, mi_row + hbs, mi_col + hbs, r, subsize);
+        decode_modes_sb(cm, xd, tile, mi_row, mi_col, r, subsize,
+                        token_cache);
+        decode_modes_sb(cm, xd, tile, mi_row, mi_col + hbs, r, subsize,
+                        token_cache);
+        decode_modes_sb(cm, xd, tile, mi_row + hbs, mi_col, r, subsize,
+                        token_cache);
+        decode_modes_sb(cm, xd, tile, mi_row + hbs, mi_col + hbs, r, subsize,
+                        token_cache);
         break;
       default:
         assert(!"Invalid partition type");
@@ -790,7 +803,8 @@
     vp9_zero(xd->left_seg_context);
     for (mi_col = tile->mi_col_start; mi_col < tile->mi_col_end;
          mi_col += MI_BLOCK_SIZE)
-      decode_modes_sb(cm, xd, tile, mi_row, mi_col, r, BLOCK_64X64);
+      decode_modes_sb(cm, xd, tile, mi_row, mi_col, r, BLOCK_64X64,
+                      pbi->token_cache);
 
     if (pbi->do_loopfilter_inline) {
       const int lf_start = mi_row - MI_BLOCK_SIZE;
@@ -934,7 +948,7 @@
 }
 
 static int tile_worker_hook(void *arg1, void *arg2) {
-  TileWorkerData *const tile_data = (TileWorkerData*)arg1;
+  TileWorkerData *tile_data = (TileWorkerData*)arg1;
   const TileInfo *const tile = (TileInfo*)arg2;
   int mi_row, mi_col;
 
@@ -943,9 +957,11 @@
     vp9_zero(tile_data->xd.left_context);
     vp9_zero(tile_data->xd.left_seg_context);
     for (mi_col = tile->mi_col_start; mi_col < tile->mi_col_end;
-         mi_col += MI_BLOCK_SIZE)
+         mi_col += MI_BLOCK_SIZE) {
       decode_modes_sb(tile_data->cm, &tile_data->xd, tile,
-                      mi_row, mi_col, &tile_data->bit_reader, BLOCK_64X64);
+                      mi_row, mi_col, &tile_data->bit_reader, BLOCK_64X64,
+                      tile_data->token_cache);
+    }
   }
   return !tile_data->xd.corrupted;
 }
@@ -1209,7 +1225,7 @@
 
     for (j = 0; j < PARTITION_CONTEXTS; ++j)
       for (i = 0; i < PARTITION_TYPES - 1; ++i)
-        vp9_diff_update_prob(&r, &fc->partition_prob[INTER_FRAME][j][i]);
+        vp9_diff_update_prob(&r, &fc->partition_prob[j][i]);
 
     read_mv_probs(&r, nmvc, cm->allow_high_precision_mv);
   }
diff --git a/vp9/decoder/vp9_detokenize.c b/vp9/decoder/vp9_detokenize.c
index 6ecce28..70d0d74 100644
--- a/vp9/decoder/vp9_detokenize.c
+++ b/vp9/decoder/vp9_detokenize.c
@@ -70,7 +70,6 @@
                      DCT_EOB_MODEL_TOKEN : TWO_TOKEN) :  \
                     token];                              \
     }                                                    \
-    token_cache[scan[c]] = vp9_pt_energy_class[token];   \
   } while (0)
 
 #define WRITE_COEF_CONTINUE(val, token)                  \
@@ -78,6 +77,7 @@
     qcoeff_ptr[scan[c]] = vp9_read_and_apply_sign(r, val) * \
                             dq[c > 0] / (1 + (tx_size == TX_32X32)); \
     INCREMENT_COUNT(token);                              \
+    token_cache[scan[c]] = vp9_pt_energy_class[token];   \
     c++;                                                 \
     continue;                                            \
   }
@@ -91,7 +91,8 @@
 static int decode_coefs(VP9_COMMON *cm, const MACROBLOCKD *xd,
                         vp9_reader *r, int block_idx,
                         PLANE_TYPE type, int seg_eob, int16_t *qcoeff_ptr,
-                        TX_SIZE tx_size, const int16_t *dq, int pt) {
+                        TX_SIZE tx_size, const int16_t *dq, int pt,
+                        uint8_t *token_cache) {
   const FRAME_CONTEXT *const fc = &cm->fc;
   FRAME_COUNTS *const counts = &cm->counts;
   const int ref = is_inter_block(&xd->mi_8x8[0]->mbmi);
@@ -104,7 +105,6 @@
   vp9_coeff_count_model *coef_counts = counts->coef[tx_size];
   const int16_t *scan, *nb;
   const uint8_t *const band_translate = get_band_translate(tx_size);
-  uint8_t token_cache[1024];
   get_scan(xd, tx_size, type, block_idx, &scan, &nb);
 
   while (1) {
@@ -131,6 +131,7 @@
 
     if (!vp9_read(r, prob[ZERO_CONTEXT_NODE])) {
       INCREMENT_COUNT(ZERO_TOKEN);
+      token_cache[scan[c]] = vp9_pt_energy_class[ZERO_TOKEN];
       ++c;
       goto SKIP_START;
     }
@@ -212,7 +213,8 @@
 
 int vp9_decode_block_tokens(VP9_COMMON *cm, MACROBLOCKD *xd,
                             int plane, int block, BLOCK_SIZE plane_bsize,
-                            TX_SIZE tx_size, vp9_reader *r) {
+                            TX_SIZE tx_size, vp9_reader *r,
+                            uint8_t *token_cache) {
   struct macroblockd_plane *const pd = &xd->plane[plane];
   const int seg_eob = get_tx_eob(&cm->seg, xd->mi_8x8[0]->mbmi.segment_id,
                                  tx_size);
@@ -223,7 +225,7 @@
 
   eob = decode_coefs(cm, xd, r, block,
                      pd->plane_type, seg_eob, BLOCK_OFFSET(pd->qcoeff, block),
-                     tx_size, pd->dequant, pt);
+                     tx_size, pd->dequant, pt, token_cache);
 
   set_contexts(xd, pd, plane_bsize, tx_size, eob > 0, aoff, loff);
 
diff --git a/vp9/decoder/vp9_detokenize.h b/vp9/decoder/vp9_detokenize.h
index 94dd8e4..04939ea 100644
--- a/vp9/decoder/vp9_detokenize.h
+++ b/vp9/decoder/vp9_detokenize.h
@@ -17,6 +17,7 @@
 
 int vp9_decode_block_tokens(VP9_COMMON *cm, MACROBLOCKD *xd,
                             int plane, int block, BLOCK_SIZE plane_bsize,
-                            TX_SIZE tx_size, vp9_reader *r);
+                            TX_SIZE tx_size, vp9_reader *r,
+                            uint8_t *token_cache);
 
 #endif  // VP9_DECODER_VP9_DETOKENIZE_H_
diff --git a/vp9/decoder/vp9_onyxd_int.h b/vp9/decoder/vp9_onyxd_int.h
index 83ea967..7c4c9db 100644
--- a/vp9/decoder/vp9_onyxd_int.h
+++ b/vp9/decoder/vp9_onyxd_int.h
@@ -49,6 +49,8 @@
 
   ENTROPY_CONTEXT *above_context[MAX_MB_PLANE];
   PARTITION_CONTEXT *above_seg_context;
+
+  DECLARE_ALIGNED(16, unsigned char, token_cache[1024]);
 } VP9D_COMP;
 
 #endif  // VP9_DECODER_VP9_ONYXD_INT_H_
diff --git a/vp9/encoder/vp9_bitstream.c b/vp9/encoder/vp9_bitstream.c
index a996e0e..87bd36c 100644
--- a/vp9/encoder/vp9_bitstream.c
+++ b/vp9/encoder/vp9_bitstream.c
@@ -163,18 +163,13 @@
   vp9_wb_write_literal(wb, data, get_unsigned_bits(max));
 }
 
-static void update_mode(
-  vp9_writer *w,
-  int n,
-  vp9_tree tree,
-  vp9_prob Pnew[/* n-1 */],
-  vp9_prob Pcur[/* n-1 */],
-  unsigned int bct[/* n-1 */] [2],
-  const unsigned int num_events[/* n */]
-) {
+static void update_mode(vp9_writer *w, int n, vp9_tree tree,
+                        vp9_prob Pcur[/* n-1 */],
+                        unsigned int bct[/* n-1 */][2],
+                        const unsigned int num_events[/* n */]) {
   int i = 0;
 
-  vp9_tree_probs_from_distribution(tree, Pnew, bct, num_events, 0);
+  vp9_tree_probs_from_distribution(tree, bct, num_events, 0);
   n--;
 
   for (i = 0; i < n; ++i)
@@ -185,11 +180,10 @@
                                       vp9_writer* const bc) {
   VP9_COMMON *const cm = &cpi->common;
   int j;
-  vp9_prob pnew[INTRA_MODES - 1];
   unsigned int bct[INTRA_MODES - 1][2];
 
   for (j = 0; j < BLOCK_SIZE_GROUPS; j++)
-    update_mode(bc, INTRA_MODES, vp9_intra_mode_tree, pnew,
+    update_mode(bc, INTRA_MODES, vp9_intra_mode_tree,
                 cm->fc.y_mode_prob[j], bct,
                 (unsigned int *)cpi->y_mode_count[j]);
 }
@@ -231,43 +225,35 @@
   write_token(bc, vp9_intra_mode_tree, p, vp9_intra_mode_encodings + m);
 }
 
-static void update_switchable_interp_probs(VP9_COMP *const cpi,
-                                           vp9_writer* const bc) {
+static void update_switchable_interp_probs(VP9_COMP *cpi, vp9_writer *w) {
   VP9_COMMON *const cm = &cpi->common;
-  unsigned int branch_ct[SWITCHABLE_FILTER_CONTEXTS][SWITCHABLE_FILTERS - 1][2];
-  vp9_prob new_prob[SWITCHABLE_FILTER_CONTEXTS][SWITCHABLE_FILTERS - 1];
+  unsigned int branch_ct[SWITCHABLE_FILTERS - 1][2];
   int i, j;
   for (j = 0; j < SWITCHABLE_FILTER_CONTEXTS; ++j) {
-    vp9_tree_probs_from_distribution(
-        vp9_switchable_interp_tree,
-        new_prob[j], branch_ct[j],
-        cm->counts.switchable_interp[j], 0);
+    vp9_tree_probs_from_distribution(vp9_switchable_interp_tree, branch_ct,
+                                     cm->counts.switchable_interp[j], 0);
+
+    for (i = 0; i < SWITCHABLE_FILTERS - 1; ++i)
+      vp9_cond_prob_diff_update(w, &cm->fc.switchable_interp_prob[j][i],
+                                branch_ct[i]);
   }
-  for (j = 0; j < SWITCHABLE_FILTER_CONTEXTS; ++j) {
-    for (i = 0; i < SWITCHABLE_FILTERS - 1; ++i) {
-      vp9_cond_prob_diff_update(bc, &cm->fc.switchable_interp_prob[j][i],
-                                branch_ct[j][i]);
-    }
-  }
+
 #ifdef MODE_STATS
   if (!cpi->dummy_packing)
     update_switchable_interp_stats(cm);
 #endif
 }
 
-static void update_inter_mode_probs(VP9_COMMON *cm, vp9_writer* const bc) {
+static void update_inter_mode_probs(VP9_COMMON *cm, vp9_writer *w) {
   int i, j;
 
   for (i = 0; i < INTER_MODE_CONTEXTS; ++i) {
     unsigned int branch_ct[INTER_MODES - 1][2];
-    vp9_prob new_prob[INTER_MODES - 1];
-
-    vp9_tree_probs_from_distribution(vp9_inter_mode_tree,
-                                     new_prob, branch_ct,
+    vp9_tree_probs_from_distribution(vp9_inter_mode_tree, branch_ct,
                                      cm->counts.inter_mode[i], NEARESTMV);
 
     for (j = 0; j < INTER_MODES - 1; ++j)
-      vp9_cond_prob_diff_update(bc, &cm->fc.inter_mode_probs[i][j],
+      vp9_cond_prob_diff_update(w, &cm->fc.inter_mode_probs[i][j],
                                 branch_ct[j]);
   }
 }
@@ -592,25 +578,26 @@
   pack_mb_tokens(bc, tok, tok_end);
 }
 
-static void write_partition(PARTITION_TYPE partition,
-                            int hbs, int mi_rows, int mi_cols,
-                            int mi_row, int mi_col,
-                            vp9_prob probs[PARTITION_TYPES - 1],
-                            vp9_writer *w) {
-  const int has_rows = (mi_row + hbs) < mi_rows;
-  const int has_cols = (mi_col + hbs) < mi_cols;
+static void write_partition(VP9_COMP *cpi, int hbs, int mi_row, int mi_col,
+                            PARTITION_TYPE p, BLOCK_SIZE bsize, vp9_writer *w) {
+  VP9_COMMON *const cm = &cpi->common;
+  const int ctx = partition_plane_context(cpi->above_seg_context,
+                                          cpi->left_seg_context,
+                                          mi_row, mi_col, bsize);
+  const vp9_prob *const probs = get_partition_probs(cm, ctx);
+  const int has_rows = (mi_row + hbs) < cm->mi_rows;
+  const int has_cols = (mi_col + hbs) < cm->mi_cols;
 
   if (has_rows && has_cols) {
-    write_token(w, vp9_partition_tree, probs,
-                &vp9_partition_encodings[partition]);
+    write_token(w, vp9_partition_tree, probs, &vp9_partition_encodings[p]);
   } else if (!has_rows && has_cols) {
-    assert(partition == PARTITION_SPLIT || partition == PARTITION_HORZ);
-    vp9_write(w, partition == PARTITION_SPLIT, probs[1]);
+    assert(p == PARTITION_SPLIT || p == PARTITION_HORZ);
+    vp9_write(w, p == PARTITION_SPLIT, probs[1]);
   } else if (has_rows && !has_cols) {
-    assert(partition == PARTITION_SPLIT || partition == PARTITION_VERT);
-    vp9_write(w, partition == PARTITION_SPLIT, probs[2]);
+    assert(p == PARTITION_SPLIT || p == PARTITION_VERT);
+    vp9_write(w, p == PARTITION_SPLIT, probs[2]);
   } else {
-    assert(partition == PARTITION_SPLIT);
+    assert(p == PARTITION_SPLIT);
   }
 }
 
@@ -637,11 +624,7 @@
     if (index > 0)
       return;
   } else {
-    const int ctx = partition_plane_context(cpi->above_seg_context,
-                                            cpi->left_seg_context,
-                                            mi_row, mi_col, bsize);
-    write_partition(partition, bs, cm->mi_rows, cm->mi_cols, mi_row, mi_col,
-                    cm->fc.partition_prob[cm->frame_type][ctx], bc);
+    write_partition(cpi, bs, mi_row, mi_col, partition, bsize, bc);
   }
 
   subsize = get_subsize(bsize, partition);
@@ -710,8 +693,7 @@
   unsigned int (*eob_branch_ct)[REF_TYPES][COEF_BANDS][PREV_COEF_CONTEXTS] =
       cpi->common.counts.eob_branch[tx_size];
   vp9_coeff_stats *coef_branch_ct = cpi->frame_branch_ct[tx_size];
-  vp9_prob full_probs[ENTROPY_NODES];
-  int i, j, k, l;
+  int i, j, k, l, m;
 
   for (i = 0; i < BLOCK_TYPES; ++i) {
     for (j = 0; j < REF_TYPES; ++j) {
@@ -720,16 +702,14 @@
           if (l >= 3 && k == 0)
             continue;
           vp9_tree_probs_from_distribution(vp9_coef_tree,
-                                           full_probs,
                                            coef_branch_ct[i][j][k][l],
                                            coef_counts[i][j][k][l], 0);
-          vpx_memcpy(coef_probs[i][j][k][l], full_probs,
-                     sizeof(vp9_prob) * UNCONSTRAINED_NODES);
           coef_branch_ct[i][j][k][l][0][1] = eob_branch_ct[i][j][k][l] -
                                              coef_branch_ct[i][j][k][l][0][0];
-          coef_probs[i][j][k][l][0] =
-              get_binary_prob(coef_branch_ct[i][j][k][l][0][0],
-                              coef_branch_ct[i][j][k][l][0][1]);
+          for (m = 0; m < UNCONSTRAINED_NODES; ++m)
+            coef_probs[i][j][k][l][m] = get_binary_prob(
+                                            coef_branch_ct[i][j][k][l][m][0],
+                                            coef_branch_ct[i][j][k][l][m][1]);
 #ifdef ENTROPY_STATS
           if (!cpi->dummy_packing) {
             int t;
@@ -1467,11 +1447,9 @@
     update_mbintra_mode_probs(cpi, &header_bc);
 
     for (i = 0; i < PARTITION_CONTEXTS; ++i) {
-      vp9_prob pnew[PARTITION_TYPES - 1];
       unsigned int bct[PARTITION_TYPES - 1][2];
-      update_mode(&header_bc, PARTITION_TYPES,
-                  vp9_partition_tree, pnew,
-                  fc->partition_prob[cm->frame_type][i], bct,
+      update_mode(&header_bc, PARTITION_TYPES, vp9_partition_tree,
+                  fc->partition_prob[i], bct,
                   (unsigned int *)cpi->partition_count[i]);
     }
 
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index 44ade18..86332bc 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -683,10 +683,6 @@
                                [mbmi->ref_frame[0] != GOLDEN_FRAME]++;
       }
     }
-
-    // Count of last ref frame 0,0 usage
-    if (mbmi->mode == ZEROMV && mbmi->ref_frame[0] == LAST_FRAME)
-      cpi->inter_zz_count++;
   }
 }
 
@@ -1931,9 +1927,6 @@
 
   totalrate = 0;
 
-  // Reset frame count of inter 0,0 motion vector usage.
-  cpi->inter_zz_count = 0;
-
   vp9_zero(cm->counts.switchable_interp);
   vp9_zero(cpi->tx_stepdown_count);
 
diff --git a/vp9/encoder/vp9_encodemv.c b/vp9/encoder/vp9_encodemv.c
index 9ebcc49..e2c6c4c 100644
--- a/vp9/encoder/vp9_encodemv.c
+++ b/vp9/encoder/vp9_encodemv.c
@@ -124,8 +124,9 @@
   }
 }
 
-static int update_mv(vp9_writer *w, const unsigned int ct[2],
-                     vp9_prob *cur_p, vp9_prob new_p, vp9_prob upd_p) {
+static int update_mv(vp9_writer *w, const unsigned int ct[2], vp9_prob *cur_p,
+                     vp9_prob upd_p) {
+  const vp9_prob new_p = get_binary_prob(ct[0], ct[1]);
   vp9_prob mod_p = new_p | 1;
   const int cur_b = cost_branch256(ct, *cur_p);
   const int mod_b = cost_branch256(ct, mod_p);
@@ -143,7 +144,6 @@
 
 static void counts_to_nmv_context(
     nmv_context_counts *nmv_count,
-    nmv_context *prob,
     int usehp,
     unsigned int (*branch_ct_joint)[2],
     unsigned int (*branch_ct_sign)[2],
@@ -156,29 +156,24 @@
     unsigned int (*branch_ct_hp)[2]) {
   int i, j, k;
   vp9_tree_probs_from_distribution(vp9_mv_joint_tree,
-                                   prob->joints,
                                    branch_ct_joint,
                                    nmv_count->joints, 0);
   for (i = 0; i < 2; ++i) {
     const uint32_t s0 = nmv_count->comps[i].sign[0];
     const uint32_t s1 = nmv_count->comps[i].sign[1];
 
-    prob->comps[i].sign = get_binary_prob(s0, s1);
     branch_ct_sign[i][0] = s0;
     branch_ct_sign[i][1] = s1;
     vp9_tree_probs_from_distribution(vp9_mv_class_tree,
-                                     prob->comps[i].classes,
-                                     branch_ct_classes[i],
-                                     nmv_count->comps[i].classes, 0);
+                                    branch_ct_classes[i],
+                                    nmv_count->comps[i].classes, 0);
     vp9_tree_probs_from_distribution(vp9_mv_class0_tree,
-                                     prob->comps[i].class0,
                                      branch_ct_class0[i],
                                      nmv_count->comps[i].class0, 0);
     for (j = 0; j < MV_OFFSET_BITS; ++j) {
       const uint32_t b0 = nmv_count->comps[i].bits[j][0];
       const uint32_t b1 = nmv_count->comps[i].bits[j][1];
 
-      prob->comps[i].bits[j] = get_binary_prob(b0, b1);
       branch_ct_bits[i][j][0] = b0;
       branch_ct_bits[i][j][1] = b1;
     }
@@ -186,12 +181,10 @@
   for (i = 0; i < 2; ++i) {
     for (k = 0; k < CLASS0_SIZE; ++k) {
       vp9_tree_probs_from_distribution(vp9_mv_fp_tree,
-                                       prob->comps[i].class0_fp[k],
                                        branch_ct_class0_fp[i][k],
                                        nmv_count->comps[i].class0_fp[k], 0);
     }
     vp9_tree_probs_from_distribution(vp9_mv_fp_tree,
-                                     prob->comps[i].fp,
                                      branch_ct_fp[i],
                                      nmv_count->comps[i].fp, 0);
   }
@@ -202,11 +195,9 @@
       const uint32_t hp0 = nmv_count->comps[i].hp[0];
       const uint32_t hp1 = nmv_count->comps[i].hp[1];
 
-      prob->comps[i].class0_hp = get_binary_prob(c0_hp0, c0_hp1);
       branch_ct_class0_hp[i][0] = c0_hp0;
       branch_ct_class0_hp[i][1] = c0_hp1;
 
-      prob->comps[i].hp = get_binary_prob(hp0, hp1);
       branch_ct_hp[i][0] = hp0;
       branch_ct_hp[i][1] = hp1;
     }
@@ -215,7 +206,6 @@
 
 void vp9_write_nmv_probs(VP9_COMP* const cpi, int usehp, vp9_writer* const bc) {
   int i, j;
-  nmv_context prob;
   unsigned int branch_ct_joint[MV_JOINTS - 1][2];
   unsigned int branch_ct_sign[2][2];
   unsigned int branch_ct_classes[2][MV_CLASSES - 1][2];
@@ -227,30 +217,28 @@
   unsigned int branch_ct_hp[2][2];
   nmv_context *mvc = &cpi->common.fc.nmvc;
 
-  counts_to_nmv_context(&cpi->NMVcount, &prob, usehp,
+  counts_to_nmv_context(&cpi->NMVcount, usehp,
                         branch_ct_joint, branch_ct_sign, branch_ct_classes,
                         branch_ct_class0, branch_ct_bits,
                         branch_ct_class0_fp, branch_ct_fp,
                         branch_ct_class0_hp, branch_ct_hp);
 
   for (j = 0; j < MV_JOINTS - 1; ++j)
-    update_mv(bc, branch_ct_joint[j], &mvc->joints[j], prob.joints[j],
-              NMV_UPDATE_PROB);
+    update_mv(bc, branch_ct_joint[j], &mvc->joints[j], NMV_UPDATE_PROB);
 
   for (i = 0; i < 2; ++i) {
-    update_mv(bc, branch_ct_sign[i], &mvc->comps[i].sign,
-              prob.comps[i].sign, NMV_UPDATE_PROB);
+    update_mv(bc, branch_ct_sign[i], &mvc->comps[i].sign, NMV_UPDATE_PROB);
     for (j = 0; j < MV_CLASSES - 1; ++j)
       update_mv(bc, branch_ct_classes[i][j], &mvc->comps[i].classes[j],
-                prob.comps[i].classes[j], NMV_UPDATE_PROB);
+                NMV_UPDATE_PROB);
 
     for (j = 0; j < CLASS0_SIZE - 1; ++j)
       update_mv(bc, branch_ct_class0[i][j], &mvc->comps[i].class0[j],
-                prob.comps[i].class0[j], NMV_UPDATE_PROB);
+                NMV_UPDATE_PROB);
 
     for (j = 0; j < MV_OFFSET_BITS; ++j)
       update_mv(bc, branch_ct_bits[i][j], &mvc->comps[i].bits[j],
-                prob.comps[i].bits[j], NMV_UPDATE_PROB);
+                NMV_UPDATE_PROB);
   }
 
   for (i = 0; i < 2; ++i) {
@@ -258,21 +246,19 @@
       int k;
       for (k = 0; k < 3; ++k)
         update_mv(bc, branch_ct_class0_fp[i][j][k],
-                  &mvc->comps[i].class0_fp[j][k],
-                  prob.comps[i].class0_fp[j][k], NMV_UPDATE_PROB);
+                  &mvc->comps[i].class0_fp[j][k], NMV_UPDATE_PROB);
     }
 
     for (j = 0; j < 3; ++j)
-      update_mv(bc, branch_ct_fp[i][j], &mvc->comps[i].fp[j],
-                prob.comps[i].fp[j], NMV_UPDATE_PROB);
+      update_mv(bc, branch_ct_fp[i][j], &mvc->comps[i].fp[j], NMV_UPDATE_PROB);
   }
 
   if (usehp) {
     for (i = 0; i < 2; ++i) {
       update_mv(bc, branch_ct_class0_hp[i], &mvc->comps[i].class0_hp,
-                prob.comps[i].class0_hp, NMV_UPDATE_PROB);
+                NMV_UPDATE_PROB);
       update_mv(bc, branch_ct_hp[i], &mvc->comps[i].hp,
-                prob.comps[i].hp, NMV_UPDATE_PROB);
+                NMV_UPDATE_PROB);
     }
   }
 }
diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c
index c83954e..c3dbc86 100644
--- a/vp9/encoder/vp9_firstpass.c
+++ b/vp9/encoder/vp9_firstpass.c
@@ -2169,17 +2169,14 @@
       cpi->ni_av_qi = tmp_q;
       cpi->avg_q = vp9_convert_qindex_to_q(tmp_q);
 
-#ifndef ONE_SHOT_Q_ESTIMATE
       // Limit the maxq value returned subsequently.
       // This increases the risk of overspend or underspend if the initial
       // estimate for the clip is bad, but helps prevent excessive
       // variation in Q, especially near the end of a clip
       // where for example a small overspend may cause Q to crash
       adjust_maxq_qrange(cpi);
-#endif
     }
 
-#ifndef ONE_SHOT_Q_ESTIMATE
     // The last few frames of a clip almost always have to few or too many
     // bits and for the sake of over exact rate control we dont want to make
     // radical adjustments to the allowed quantizer range just to use up a
@@ -2202,7 +2199,6 @@
       cpi->active_worst_quality =
           adjust_active_maxq(cpi->active_worst_quality, tmp_q);
     }
-#endif
   }
   vp9_zero(this_frame);
   if (EOF == input_stats(cpi, &this_frame))
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index b664f1e..1d3170a 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -1032,11 +1032,6 @@
     CHECK_MEM_ERROR(cm, cpi->tok, vpx_calloc(tokens, sizeof(*cpi->tok)));
   }
 
-  // Data used for real time vc mode to see if gf needs refreshing
-  cpi->inter_zz_count = 0;
-  cpi->gf_bad_count = 0;
-  cpi->gf_update_recommended = 0;
-
   vpx_free(cpi->mb_activity_map);
   CHECK_MEM_ERROR(cm, cpi->mb_activity_map,
                   vpx_calloc(sizeof(unsigned int),
@@ -1185,7 +1180,6 @@
   int i;
 
   cpi->oxcf = *oxcf;
-  cpi->goldfreq = 7;
 
   cm->version = oxcf->version;
 
@@ -2856,19 +2850,11 @@
     if (cpi->oxcf.end_usage == USAGE_CONSTANT_QUALITY) {
       cpi->active_best_quality = cpi->cq_target_quality;
     } else {
-#ifdef ONE_SHOT_Q_ESTIMATE
-#ifdef STRICT_ONE_SHOT_Q
-      cpi->active_best_quality = q;
-#else
-      cpi->active_best_quality = inter_minq[q];
-#endif
-#else
       cpi->active_best_quality = inter_minq[q];
       // 1-pass: for now, use the average Q for the active_best, if its lower
       // than active_worst.
       if (cpi->pass == 0 && (cpi->avg_frame_qindex < q))
         cpi->active_best_quality = inter_minq[cpi->avg_frame_qindex];
-#endif
 
       // For the constrained quality mode we don't want
       // q to fall below the cq level.
diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index 0498043..9429c7f 100644
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -29,11 +29,6 @@
 #include "vp9/common/vp9_findnearmv.h"
 #include "vp9/encoder/vp9_lookahead.h"
 
-// Experimental rate control switches
-#if CONFIG_ONESHOTQ
-#define ONE_SHOT_Q_ESTIMATE 0
-#define STRICT_ONE_SHOT_Q 0
-#endif
 #define DISABLE_RC_LONG_TERM_MEM 0
 
 // #define MODE_TEST_HIT_STATS
@@ -506,14 +501,9 @@
   int decimation_count;
 
   // for real time encoding
-  int avg_encode_time;              // microsecond
-  int avg_pick_mode_time;            // microsecond
   int speed;
-  unsigned int cpu_freq;           // Mhz
   int compressor_speed;
 
-  int interquantizer;
-  int goldfreq;
   int auto_worst_q;
   int cpu_used;
   int pass;
@@ -529,12 +519,6 @@
   unsigned int max_mv_magnitude;
   int mv_step_param;
 
-  // Data used for real time conferencing mode to help determine if it
-  // would be good to update the gf
-  int inter_zz_count;
-  int gf_bad_count;
-  int gf_update_recommended;
-
   unsigned char *segmentation_map;
 
   // segment threashold for encode breakout
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index f69b19b..993919e 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -251,8 +251,7 @@
   fill_token_costs(cpi->mb.token_costs, cm->fc.coef_probs);
 
   for (i = 0; i < PARTITION_CONTEXTS; i++)
-    vp9_cost_tokens(cpi->mb.partition_cost[i],
-                    cm->fc.partition_prob[cm->frame_type][i],
+    vp9_cost_tokens(cpi->mb.partition_cost[i], get_partition_probs(cm, i),
                     vp9_partition_tree);
 
   /*rough estimate for costing*/
@@ -1093,7 +1092,7 @@
         else
           x->fwd_txm4x4(src_diff, coeff, 8);
 
-        vp9_regular_quantize_b_4x4(x, 16, block, scan, get_iscan_4x4(tx_type));
+        vp9_regular_quantize_b_4x4(x, 4, block, scan, get_iscan_4x4(tx_type));
 
         ratey += cost_coeffs(x, 0, block,
                              tempa + idx, templ + idy, TX_4X4, scan, nb);
@@ -1560,7 +1559,7 @@
       coeff = BLOCK_OFFSET(p->coeff, k);
       x->fwd_txm4x4(raster_block_offset_int16(BLOCK_8X8, k, p->src_diff),
                     coeff, 8);
-      vp9_regular_quantize_b_4x4(x, 16, k, get_scan_4x4(DCT_DCT),
+      vp9_regular_quantize_b_4x4(x, 4, k, get_scan_4x4(DCT_DCT),
                                  get_iscan_4x4(DCT_DCT));
       thisdistortion += vp9_block_error(coeff, BLOCK_OFFSET(pd->dqcoeff, k),
                                         16, &ssz);
@@ -1872,12 +1871,14 @@
           mi_buf_restore(x, orig_src, orig_pre);
         }
 
-        if (has_second_rf && this_mode == NEWMV &&
-            mbmi->interp_filter == EIGHTTAP) {
+        if (has_second_rf) {
           if (seg_mvs[i][mbmi->ref_frame[1]].as_int == INVALID_MV ||
               seg_mvs[i][mbmi->ref_frame[0]].as_int == INVALID_MV)
             continue;
+        }
 
+        if (has_second_rf && this_mode == NEWMV &&
+            mbmi->interp_filter == EIGHTTAP) {
           // adjust src pointers
           mi_buf_shift(x, i);
           if (cpi->sf.comp_inter_joint_search_thresh <= bsize) {
@@ -2661,6 +2662,12 @@
   int orig_dst_stride[MAX_MB_PLANE];
   int rs = 0;
 
+  if (is_comp_pred) {
+    if (frame_mv[refs[0]].as_int == INVALID_MV ||
+        frame_mv[refs[1]].as_int == INVALID_MV)
+      return INT64_MAX;
+  }
+
   if (this_mode == NEWMV) {
     int rate_mv;
     if (is_comp_pred) {
@@ -2679,9 +2686,6 @@
                                    &mbmi->ref_mvs[refs[1]][0].as_mv,
                                    x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
       }
-      if (frame_mv[refs[0]].as_int == INVALID_MV ||
-          frame_mv[refs[1]].as_int == INVALID_MV)
-        return INT64_MAX;
       *rate2 += rate_mv;
     } else {
       int_mv tmp_mv;
@@ -3544,8 +3548,9 @@
     }
 
     // Keep record of best inter rd with single reference
-    if (is_inter_block(&xd->mi_8x8[0]->mbmi) && !mode_excluded &&
-        this_rd < best_inter_rd) {
+    if (is_inter_block(&xd->mi_8x8[0]->mbmi) &&
+        !has_second_ref(&xd->mi_8x8[0]->mbmi) &&
+        !mode_excluded && this_rd < best_inter_rd) {
       best_inter_rd = this_rd;
       best_inter_ref_frame = ref_frame;
     }
diff --git a/vp9/encoder/vp9_subexp.c b/vp9/encoder/vp9_subexp.c
index eb864d9..387fc90 100644
--- a/vp9/encoder/vp9_subexp.c
+++ b/vp9/encoder/vp9_subexp.c
@@ -221,7 +221,7 @@
 }
 
 void vp9_cond_prob_diff_update(vp9_writer *w, vp9_prob *oldp,
-                               unsigned int *ct) {
+                               const unsigned int ct[2]) {
   const vp9_prob upd = DIFF_UPDATE_PROB;
   vp9_prob newp = get_binary_prob(ct[0], ct[1]);
   const int savings = vp9_prob_diff_update_savings_search(ct, *oldp, &newp,