New inter mode context.

This patch creates a new inter mode contest that avoids
a dependence on the reconstructed motion vectors from
neighboring blocks. This was a change requested by
a hardware vendor to improve decode performance.

As part of this change I have also made some modifications
to stats output code (under a flag) to allow accumulation of
inter mode context flags over multiple clips

Some further changes will be required to accommodate the
deprecation of the split mv mode over the next few days.

Performance as stands is around -0.25% on derf and
std-hd but up on the YT and YT-HD sets. With further tuning
or some adjustment to the context criteria it should be
possible to make this change broadly neutral.

Change-Id: Ia15cb4470969b9e87332a59c546ae0bd40676f6c

5 files changed, 71 insertions, 36 deletions

diff --git a/vp9/encoder/vp9_bitstream.c b/vp9/encoder/vp9_bitstream.c
index 398545136..fbdee35f9 100644
--- a/vp9/encoder/vp9_bitstream.c
+++ b/vp9/encoder/vp9_bitstream.c
@@ -826,9 +826,9 @@ static void write_mb_modes_kf(const VP9_COMP *cpi,
                                   left_block_mode(m, i) : B_DC_PRED;
       const int bm = m->bmi[i].as_mode.first;
 
-#ifdef ENTROPY_STATS
+/*#ifdef ENTROPY_STATS
       ++intra_mode_stats [A] [L] [bm];
-#endif
+#endif*/
       write_kf_bmode(bc, bm, c->kf_bmode_prob[a][l]);
     } while (++i < 4);
   }
diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c
index aff5637e1..75e6e6757 100644
--- a/vp9/encoder/vp9_mcomp.c
+++ b/vp9/encoder/vp9_mcomp.c
@@ -2430,32 +2430,3 @@ int vp9_refining_search_8p_c(MACROBLOCK *x,
   }
 }
 #endif  // CONFIG_COMP_INTER_JOINT_SEARCH
-
-#ifdef ENTROPY_STATS
-void print_mode_context(VP9_COMMON *pc) {
-  FILE *f = fopen("vp9_modecont.c", "a");
-  int i, j;
-
-  fprintf(f, "#include \"vp9_entropy.h\"\n");
-  fprintf(f, "const int vp9_mode_contexts[INTER_MODE_CONTEXTS][4] =");
-  fprintf(f, "{\n");
-  for (j = 0; j < INTER_MODE_CONTEXTS; j++) {
-    fprintf(f, "  {/* %d */ ", j);
-    fprintf(f, "    ");
-    for (i = 0; i < 4; i++) {
-      int this_prob;
-
-      // context probs
-      this_prob = get_binary_prob(pc->fc.mv_ref_ct[j][i][0],
-                                  pc->fc.mv_ref_ct[j][i][1]);
-
-      fprintf(f, "%5d, ", this_prob);
-    }
-    fprintf(f, "  },\n");
-  }
-
-  fprintf(f, "};\n");
-  fclose(f);
-}
-
-#endif/* END MV ref count ENTROPY_STATS stats code */
diff --git a/vp9/encoder/vp9_mcomp.h b/vp9/encoder/vp9_mcomp.h
index cdbd29aa5..33e688b97 100644
--- a/vp9/encoder/vp9_mcomp.h
+++ b/vp9/encoder/vp9_mcomp.h
@@ -15,10 +15,6 @@
 #include "vp9/encoder/vp9_block.h"
 #include "vp9/encoder/vp9_variance.h"
 
-#ifdef ENTROPY_STATS
-void print_mode_context(VP9_COMMON *pc);
-#endif
-
 // The maximum number of steps in a step search given the largest
 // allowed initial step
 #define MAX_MVSEARCH_STEPS 11
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index 67d1b67fc..86d32ea2f 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -533,6 +533,66 @@ static void configure_implicit_segmentation(VP9_COMP *cpi, int frame_qindex) {
 }
 #endif
 
+#ifdef ENTROPY_STATS
+void vp9_update_mode_context_stats(VP9_COMP *cpi) {
+  VP9_COMMON *cm = &cpi->common;
+  int i, j;
+  unsigned int (*mv_ref_ct)[4][2] = cm->fc.mv_ref_ct;
+  int64_t (*mv_ref_stats)[4][2] = cpi->mv_ref_stats;
+  FILE *f;
+
+  // Read the past stats counters
+  f = fopen("mode_context.bin",  "rb");
+  if (!f) {
+    vpx_memset(cpi->mv_ref_stats, 0, sizeof(cpi->mv_ref_stats));
+  } else {
+    fread(cpi->mv_ref_stats, sizeof(cpi->mv_ref_stats), 1, f);
+    fclose(f);
+  }
+
+  // Add in the values for this frame
+  for (i = 0; i < INTER_MODE_CONTEXTS; i++) {
+    for (j = 0; j < 4; j++) {
+      mv_ref_stats[i][j][0] += (int64_t)mv_ref_ct[i][j][0];
+      mv_ref_stats[i][j][1] += (int64_t)mv_ref_ct[i][j][1];
+    }
+  }
+
+  // Write back the accumulated stats
+  f = fopen("mode_context.bin",  "wb");
+  fwrite(cpi->mv_ref_stats, sizeof(cpi->mv_ref_stats), 1, f);
+  fclose(f);
+}
+
+void print_mode_context(VP9_COMP *cpi) {
+  FILE *f = fopen("vp9_modecont.c", "a");
+  int i, j;
+
+  fprintf(f, "#include \"vp9_entropy.h\"\n");
+  fprintf(f, "const int vp9_mode_contexts[INTER_MODE_CONTEXTS][4] =");
+  fprintf(f, "{\n");
+  for (j = 0; j < INTER_MODE_CONTEXTS; j++) {
+    fprintf(f, "  {/* %d */ ", j);
+    fprintf(f, "    ");
+    for (i = 0; i < 4; i++) {
+      int this_prob;
+      int64_t count = cpi->mv_ref_stats[j][i][0] + cpi->mv_ref_stats[j][i][1];
+      if (count)
+        this_prob = ((cpi->mv_ref_stats[j][i][0] * 256) + (count >> 1)) / count;
+      else
+        this_prob = 128;
+
+      // context probs
+      fprintf(f, "%5d, ", this_prob);
+    }
+    fprintf(f, "  },\n");
+  }
+
+  fprintf(f, "};\n");
+  fclose(f);
+}
+#endif  // ENTROPY_STATS
+
 // DEBUG: Print out the segment id of each MB in the current frame.
 static void print_seg_map(VP9_COMP *cpi) {
   VP9_COMMON *cm = &cpi->common;
@@ -1646,7 +1706,7 @@ void vp9_remove_compressor(VP9_PTR *ptr) {
     if (cpi->pass != 1) {
       print_context_counters();
       print_tree_update_probs();
-      print_mode_context(&cpi->common);
+      print_mode_context(cpi);
     }
 #endif
 #ifdef NMV_STATS
@@ -3239,6 +3299,10 @@ static void encode_frame_to_data_rate(VP9_COMP *cpi,
     }
   }
 
+#ifdef ENTROPY_STATS
+  vp9_update_mode_context_stats(cpi);
+#endif
+
   /* Move storing frame_type out of the above loop since it is also
    * needed in motion search besides loopfilter */
   cm->last_frame_type = cm->frame_type;
diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index aba4c0e64..9a94f3be5 100644
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -621,6 +621,10 @@ typedef struct VP9_COMP {
   int this_frame_weight;
   int max_arf_level;
 #endif
+
+#ifdef ENTROPY_STATS
+  int64_t mv_ref_stats[INTER_MODE_CONTEXTS][4][2];
+#endif
 } VP9_COMP;
 
 void vp9_encode_frame(VP9_COMP *cpi);