Expanding the coefficient encoding contexts

This patch expands the set of prev contexts used for video coding
from 3 to 4.

There is a small improvement of the order of 0.08% for derf and
0.15% on the HD set. The tests were rerun after the various merges
last week. There are two columns in each test - the first are the
results with the mbskip change, and the second with expanded contexts
added on top of that.

Derf:
http://www.corp.google.com/~debargha/vp8_results/explibvpx_newentropy_expcontext.html

HD:
http://www.corp.google.com/~debargha/vp8_results/explibvpx_hd_newentropy_expcontext.html

Rebased.

Broke up 80 char lines.

Change-Id: I82d2e72d054e530cbf5ce9aa0e6d85c582965675

2 files changed, 64 insertions, 5 deletions

diff --git a/vp8/decoder/decodframe.c b/vp8/decoder/decodframe.c
index 7cc917d09..35353805b 100644
--- a/vp8/decoder/decodframe.c
+++ b/vp8/decoder/decodframe.c
@@ -803,6 +803,12 @@ static void read_coef_probs3(VP8D_COMP *pbi)
                 for (j = !i; j < COEF_BANDS; j++)
                     for (k = 0; k < PREV_COEF_CONTEXTS; k++)
                     {
+#if CONFIG_EXPANDED_COEF_CONTEXT
+                        if (k >= 3 && ((i == 0 && j == 1) ||
+                                       (i > 0 && j == 0)))
+                            continue;
+#endif
+                        {
                         vp8_prob *const p = pc->fc.coef_probs [i][j][k] + l;
                         int u = vp8_read(bc, vp8_coef_update_probs [i][j][k][l]);
                         if (u)
@@ -810,6 +816,7 @@ static void read_coef_probs3(VP8D_COMP *pbi)
                             int delp = vp8_decode_term_subexp(bc, SUBEXP_PARAM, 255);
                             *p = (vp8_prob)inv_remap_prob(delp, *p);
                         }
+                        }
                     }
             }
         }
@@ -824,6 +831,12 @@ static void read_coef_probs3(VP8D_COMP *pbi)
                     for (j = !i; j < COEF_BANDS; j++)
                         for (k = 0; k < PREV_COEF_CONTEXTS; k++)
                         {
+#if CONFIG_EXPANDED_COEF_CONTEXT
+                            if (k >= 3 && ((i == 0 && j == 1) ||
+                                           (i > 0 && j == 0)))
+                                continue;
+#endif
+                            {
                             vp8_prob *const p = pc->fc.coef_probs_8x8 [i][j][k] + l;
                             int u = vp8_read(bc, vp8_coef_update_probs_8x8 [i][j][k][l]);
                             if (u)
@@ -831,6 +844,7 @@ static void read_coef_probs3(VP8D_COMP *pbi)
                                 int delp = vp8_decode_term_subexp(bc, SUBEXP_PARAM, 255);
                                 *p = (vp8_prob)inv_remap_prob(delp, *p);
                             }
+                            }
                         }
                 }
             }
@@ -852,6 +866,12 @@ static void read_coef_probs2(VP8D_COMP *pbi)
                 for (j = !i; j < COEF_BANDS; j++)
                     for (k = 0; k < PREV_COEF_CONTEXTS; k++)
                     {
+#if CONFIG_EXPANDED_COEF_CONTEXT
+                        if (k >= 3 && ((i == 0 && j == 1) ||
+                                       (i > 0 && j == 0)))
+                            continue;
+#endif
+                        {
                         vp8_prob *const p = pc->fc.coef_probs [i][j][k] + l;
                         int u = vp8_read(bc, vp8_coef_update_probs [i][j][k][l]);
                         if (u)
@@ -859,6 +879,7 @@ static void read_coef_probs2(VP8D_COMP *pbi)
                             int delp = vp8_decode_term_subexp(bc, SUBEXP_PARAM, 255);
                             *p = (vp8_prob)inv_remap_prob(delp, *p);
                         }
+                        }
                     }
         }
     }
@@ -872,6 +893,12 @@ static void read_coef_probs2(VP8D_COMP *pbi)
                     for (j = !i; j < COEF_BANDS; j++)
                         for (k = 0; k < PREV_COEF_CONTEXTS; k++)
                         {
+#if CONFIG_EXPANDED_COEF_CONTEXT
+                            if (k >= 3 && ((i == 0 && j == 1) ||
+                                           (i > 0 && j == 0)))
+                                continue;
+#endif
+                            {
                             vp8_prob *const p = pc->fc.coef_probs_8x8 [i][j][k] + l;
 
                             int u = vp8_read(bc, vp8_coef_update_probs_8x8 [i][j][k][l]);
@@ -880,6 +907,7 @@ static void read_coef_probs2(VP8D_COMP *pbi)
                                 int delp = vp8_decode_term_subexp(bc, SUBEXP_PARAM, 255);
                                 *p = (vp8_prob)inv_remap_prob(delp, *p);
                             }
+                            }
                         }
             }
         }
@@ -904,6 +932,12 @@ static void read_coef_probs(VP8D_COMP *pbi)
             for (j = 0; j < COEF_BANDS; j++)
 #endif
                 for (k = 0; k < PREV_COEF_CONTEXTS; k++)
+                {
+#if CONFIG_EXPANDED_COEF_CONTEXT
+                    if (k >= 3 && ((i == 0 && j == 1) ||
+                                   (i > 0 && j == 0)))
+                        continue;
+#endif
                     for (l = 0; l < ENTROPY_NODES; l++)
                     {
                         vp8_prob *const p = pc->fc.coef_probs [i][j][k] + l;
@@ -920,6 +954,7 @@ static void read_coef_probs(VP8D_COMP *pbi)
 #endif
                         }
                     }
+                }
         }
     }
 
@@ -933,6 +968,12 @@ static void read_coef_probs(VP8D_COMP *pbi)
             for (j = 0; j < COEF_BANDS; j++)
 #endif
                 for (k = 0; k < PREV_COEF_CONTEXTS; k++)
+                {
+#if CONFIG_EXPANDED_COEF_CONTEXT
+                    if (k >= 3 && ((i == 0 && j == 1) ||
+                                   (i > 0 && j == 0)))
+                        continue;
+#endif
                     for (l = 0; l < ENTROPY_NODES; l++)
                     {
 
@@ -948,6 +989,7 @@ static void read_coef_probs(VP8D_COMP *pbi)
 #endif
                         }
                     }
+                }
     }
 }
 
@@ -1390,7 +1432,6 @@ int vp8_decode_frame(VP8D_COMP *pbi)
     read_coef_probs(pbi);
 #endif
 
-
     vpx_memcpy(&xd->pre, &pc->yv12_fb[pc->lst_fb_idx], sizeof(YV12_BUFFER_CONFIG));
     vpx_memcpy(&xd->dst, &pc->yv12_fb[pc->new_fb_idx], sizeof(YV12_BUFFER_CONFIG));
 
diff --git a/vp8/decoder/detokenize.c b/vp8/decoder/detokenize.c
index 106e09663..966c0f127 100644
--- a/vp8/decoder/detokenize.c
+++ b/vp8/decoder/detokenize.c
@@ -21,14 +21,24 @@
 #define BOOL_DATA UINT8
 
 #define OCB_X PREV_COEF_CONTEXTS * ENTROPY_NODES
+
+#if CONFIG_EXPANDED_COEF_CONTEXT
+DECLARE_ALIGNED(16, static const unsigned short, coef_bands_x[16]) =
+#else
 DECLARE_ALIGNED(16, static const unsigned char, coef_bands_x[16]) =
+#endif
 {
     0 * OCB_X, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X,
     6 * OCB_X, 4 * OCB_X, 5 * OCB_X, 6 * OCB_X,
     6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X,
     6 * OCB_X, 7 * OCB_X, 7 * OCB_X, 7 * OCB_X
 };
-DECLARE_ALIGNED(64, static const unsigned char, coef_bands_x_8x8[64]) = {
+#if CONFIG_EXPANDED_COEF_CONTEXT
+DECLARE_ALIGNED(16, static const unsigned short, coef_bands_x_8x8[64]) =
+#else
+DECLARE_ALIGNED(64, static const unsigned char, coef_bands_x_8x8[64]) =
+#endif
+{
   0 * OCB_X, 1 * OCB_X, 2 * OCB_X, 3 * OCB_X, 5 * OCB_X, 4 * OCB_X, 4 * OCB_X, 5 * OCB_X,
   5 * OCB_X, 3 * OCB_X, 6 * OCB_X, 3 * OCB_X, 5 * OCB_X, 4 * OCB_X, 6 * OCB_X, 6 * OCB_X,
   6 * OCB_X, 5 * OCB_X, 5 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X, 6 * OCB_X,
@@ -208,9 +218,17 @@ DECLARE_ALIGNED(16, extern const unsigned char, vp8_norm[256]);
         NORMALIZE \
     }
 
+//#define PREV_CONTEXT_INC(val) (2+((val)>2))
+//#define PREV_CONTEXT_INC(val) (vp8_prev_token_class[(val)])
+#if CONFIG_EXPANDED_COEF_CONTEXT
+#define PREV_CONTEXT_INC(val) (vp8_prev_token_class[(val)>10?10:(val)])
+#else
+#define PREV_CONTEXT_INC(val) (2)
+#endif
+
 #define DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT(val) \
     DECODE_AND_APPLYSIGN(val) \
-    Prob = coef_probs + (ENTROPY_NODES*2); \
+    Prob = coef_probs + (ENTROPY_NODES*PREV_CONTEXT_INC(val)); \
     if(c < 15){\
         qcoeff_ptr [ scan[c] ] = (INT16) v; \
         ++c; \
@@ -221,7 +239,7 @@ DECLARE_ALIGNED(16, extern const unsigned char, vp8_norm[256]);
 
 #define DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT_8x8_2(val) \
     DECODE_AND_APPLYSIGN(val) \
-    Prob = coef_probs + (ENTROPY_NODES*2); \
+    Prob = coef_probs + (ENTROPY_NODES*PREV_CONTEXT_INC(val)); \
     if(c < 3){\
         qcoeff_ptr [ scan[c] ] = (INT16) v; \
         ++c; \
@@ -230,7 +248,7 @@ DECLARE_ALIGNED(16, extern const unsigned char, vp8_norm[256]);
     goto BLOCK_FINISHED_8x8;
 #define DECODE_SIGN_WRITE_COEFF_AND_CHECK_EXIT_8x8(val) \
     DECODE_AND_APPLYSIGN(val) \
-    Prob = coef_probs + (ENTROPY_NODES*2); \
+    Prob = coef_probs + (ENTROPY_NODES*PREV_CONTEXT_INC(val)); \
     if(c < 63){\
         qcoeff_ptr [ scan[c] ] = (INT16) v; \
         ++c; \