libvpx: Pull from upstream

source/libvpx/vp9/common/vp9_thread_common.c

 /*
  *  Copyright (c) 2014 The WebM project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
 #include "./vpx_config.h"
 #include "vpx_mem/vpx_mem.h"
-#include "vp9/common/vp9_loopfilter_thread.h"
+#include "vp9/common/vp9_entropymode.h"
+#include "vp9/common/vp9_thread_common.h"
 #include "vp9/common/vp9_reconinter.h"
 
 #if CONFIG_MULTITHREAD
 static INLINE void mutex_lock(pthread_mutex_t *const mutex) {
   const int kMaxTryLocks = 4000;
   int locked = 0;
   int i;
 
   for (i = 0; i < kMaxTryLocks; ++i) {
     if (!pthread_mutex_trylock(mutex)) {
@@ skipping 121 matching lines @@
                                 VP9LfSync *lf_sync) {
   const VP9WorkerInterface *const winterface = vp9_get_worker_interface();
   // Number of superblock rows and cols
   const int sb_rows = mi_cols_aligned_to_sb(cm->mi_rows) >> MI_BLOCK_SIZE_LOG2;
   // Decoder may allocate more threads than number of tiles based on user's
   // input.
   const int tile_cols = 1 << cm->log2_tile_cols;
   const int num_workers = MIN(nworkers, tile_cols);
   int i;
 
-  if (!lf_sync->sync_range || cm->last_height != cm->height ||
+  if (!lf_sync->sync_range || sb_rows != lf_sync->rows ||
       num_workers > lf_sync->num_workers) {
     vp9_loop_filter_dealloc(lf_sync);
     vp9_loop_filter_alloc(lf_sync, cm, sb_rows, cm->width, num_workers);
   }
 
   // Initialize cur_sb_col to -1 for all SB rows.
   vpx_memset(lf_sync->cur_sb_col, -1, sizeof(*lf_sync->cur_sb_col) * sb_rows);
 
   // Set up loopfilter thread data.
   // The decoder is capping num_workers because it has been observed that using
@@ skipping 126 matching lines @@
       vpx_free(lf_sync->cond_);
     }
 #endif  // CONFIG_MULTITHREAD
     vpx_free(lf_sync->lfdata);
     vpx_free(lf_sync->cur_sb_col);
     // clear the structure as the source of this call may be a resize in which
     // case this call will be followed by an _alloc() which may fail.
     vp9_zero(*lf_sync);
   }
 }
+
+// Accumulate frame counts.
+void vp9_accumulate_frame_counts(VP9_COMMON *cm, FRAME_COUNTS *counts,
+                                 int is_dec) {
+  int i, j, k, l, m;
+
+  for (i = 0; i < BLOCK_SIZE_GROUPS; i++)
+    for (j = 0; j < INTRA_MODES; j++)
+      cm->counts.y_mode[i][j] += counts->y_mode[i][j];
+
+  for (i = 0; i < INTRA_MODES; i++)
+    for (j = 0; j < INTRA_MODES; j++)
+      cm->counts.uv_mode[i][j] += counts->uv_mode[i][j];
+
+  for (i = 0; i < PARTITION_CONTEXTS; i++)
+    for (j = 0; j < PARTITION_TYPES; j++)
+      cm->counts.partition[i][j] += counts->partition[i][j];
+
+  if (is_dec) {
+    int n;
+    for (i = 0; i < TX_SIZES; i++)
+      for (j = 0; j < PLANE_TYPES; j++)
+        for (k = 0; k < REF_TYPES; k++)
+          for (l = 0; l < COEF_BANDS; l++)
+            for (m = 0; m < COEFF_CONTEXTS; m++) {
+              cm->counts.eob_branch[i][j][k][l][m] +=
+                  counts->eob_branch[i][j][k][l][m];
+              for (n = 0; n < UNCONSTRAINED_NODES + 1; n++)
+                cm->counts.coef[i][j][k][l][m][n] +=
+                    counts->coef[i][j][k][l][m][n];
+            }
+  } else {
+    for (i = 0; i < TX_SIZES; i++)
+      for (j = 0; j < PLANE_TYPES; j++)
+        for (k = 0; k < REF_TYPES; k++)
+          for (l = 0; l < COEF_BANDS; l++)
+            for (m = 0; m < COEFF_CONTEXTS; m++)
+              cm->counts.eob_branch[i][j][k][l][m] +=
+                  counts->eob_branch[i][j][k][l][m];
+                // In the encoder, cm->counts.coef is only updated at frame
+                // level, so not need to accumulate it here.
+                // for (n = 0; n < UNCONSTRAINED_NODES + 1; n++)
+                //   cm->counts.coef[i][j][k][l][m][n] +=
+                //       counts->coef[i][j][k][l][m][n];
+  }
+
+  for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; i++)
+    for (j = 0; j < SWITCHABLE_FILTERS; j++)
+      cm->counts.switchable_interp[i][j] += counts->switchable_interp[i][j];
+
+  for (i = 0; i < INTER_MODE_CONTEXTS; i++)
+    for (j = 0; j < INTER_MODES; j++)
+      cm->counts.inter_mode[i][j] += counts->inter_mode[i][j];
+
+  for (i = 0; i < INTRA_INTER_CONTEXTS; i++)
+    for (j = 0; j < 2; j++)
+      cm->counts.intra_inter[i][j] += counts->intra_inter[i][j];
+
+  for (i = 0; i < COMP_INTER_CONTEXTS; i++)
+    for (j = 0; j < 2; j++)
+      cm->counts.comp_inter[i][j] += counts->comp_inter[i][j];
+
+  for (i = 0; i < REF_CONTEXTS; i++)
+    for (j = 0; j < 2; j++)
+      for (k = 0; k < 2; k++)
+      cm->counts.single_ref[i][j][k] += counts->single_ref[i][j][k];
+
+  for (i = 0; i < REF_CONTEXTS; i++)
+    for (j = 0; j < 2; j++)
+      cm->counts.comp_ref[i][j] += counts->comp_ref[i][j];
+
+  for (i = 0; i < TX_SIZE_CONTEXTS; i++) {
+    for (j = 0; j < TX_SIZES; j++)
+      cm->counts.tx.p32x32[i][j] += counts->tx.p32x32[i][j];
+
+    for (j = 0; j < TX_SIZES - 1; j++)
+      cm->counts.tx.p16x16[i][j] += counts->tx.p16x16[i][j];
+
+    for (j = 0; j < TX_SIZES - 2; j++)
+      cm->counts.tx.p8x8[i][j] += counts->tx.p8x8[i][j];
+  }
+
+  for (i = 0; i < SKIP_CONTEXTS; i++)
+    for (j = 0; j < 2; j++)
+      cm->counts.skip[i][j] += counts->skip[i][j];
+
+  for (i = 0; i < MV_JOINTS; i++)
+    cm->counts.mv.joints[i] += counts->mv.joints[i];
+
+  for (k = 0; k < 2; k++) {
+    nmv_component_counts *comps = &cm->counts.mv.comps[k];
+    nmv_component_counts *comps_t = &counts->mv.comps[k];
+
+    for (i = 0; i < 2; i++) {
+      comps->sign[i] += comps_t->sign[i];
+      comps->class0_hp[i] += comps_t->class0_hp[i];
+      comps->hp[i] += comps_t->hp[i];
+    }
+
+    for (i = 0; i < MV_CLASSES; i++)
+      comps->classes[i] += comps_t->classes[i];
+
+    for (i = 0; i < CLASS0_SIZE; i++) {
+      comps->class0[i] += comps_t->class0[i];
+      for (j = 0; j < MV_FP_SIZE; j++)
+        comps->class0_fp[i][j] += comps_t->class0_fp[i][j];
+    }
+
+    for (i = 0; i < MV_OFFSET_BITS; i++)
+      for (j = 0; j < 2; j++)
+        comps->bits[i][j] += comps_t->bits[i][j];
+
+    for (i = 0; i < MV_FP_SIZE; i++)
+      comps->fp[i] += comps_t->fp[i];
+  }
+}