A host of micro-optimizations and a refactor of TimeForBoundsToIntersect

I used perf to profile the compositor thread and addressed performance problems
with the top fuctions.  The result is about a 2.5x performance boost.
(~1.3ms -> ~.5ms) on List of Pokemon for UpdateTilePriorities.

Most of the optimizations are inlining small functions.  The refactor of
TimeForBoundsToIntersect also eliminates many unnecessary float-double
conversions.


BUG=172406
NOTRY=true


Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=179770

Review URL: https://chromiumcodereview.appspot.com/12084031

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@180229 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 11 insertions, 2 deletions

diff --git a/cc/tile_manager.h b/cc/tile_manager.h
index 46d374e..d537079 100644
--- a/cc/tile_manager.h
+++ b/cc/tile_manager.h
@@ -118,13 +118,22 @@ class CC_EXPORT TileManager {
   void RegisterTile(Tile* tile);
   void UnregisterTile(Tile* tile);
   void WillModifyTilePriority(
-      Tile* tile, WhichTree tree, const TilePriority& new_priority);
+      Tile* tile, WhichTree tree, const TilePriority& new_priority) {
+    // TODO(nduca): Do something smarter if reprioritization turns out to be
+    // costly.
+    ScheduleManageTiles();
+  }
 
  private:
   void SortTiles();
   void AssignGpuMemoryToTiles();
   void FreeResourcesForTile(Tile* tile);
-  void ScheduleManageTiles();
+  void ScheduleManageTiles() {
+    if (manage_tiles_pending_)
+      return;
+    client_->ScheduleManageTiles();
+    manage_tiles_pending_ = true;
+  }
   void DispatchMoreTasks();
   void GatherPixelRefsForTile(Tile* tile);
   void DispatchImageDecodeTasksForTile(Tile* tile);