Implemented new fence syncs which replaces the old sync points.

content/common/gpu/gpu_channel.cc

Index: content/common/gpu/gpu_channel.cc
diff --git a/content/common/gpu/gpu_channel.cc b/content/common/gpu/gpu_channel.cc
index 7b0593ebbc72ea7943657e20653f66aab8be3875..46a6f43bfc6f61045eb55adaaa4a7b96f8d9652c 100644
--- a/content/common/gpu/gpu_channel.cc
+++ b/content/common/gpu/gpu_channel.cc
@@ -806,6 +806,8 @@ bool GpuChannel::OnControlMessageReceived(const IPC::Message& msg) {
                         OnCreateOffscreenCommandBuffer)
     IPC_MESSAGE_HANDLER(GpuChannelMsg_DestroyCommandBuffer,
                         OnDestroyCommandBuffer)
+    IPC_MESSAGE_HANDLER(GpuChannelMsg_Nop,
+                        OnNop)

[piman, 2015/09/26 00:09:51]

You want to handle this on the IO thread (in the filter).
Once this arrives on the IO thread, you know all previous messages have arrived
too, hence you generated a valid ordering id for them, they're in the queue,
etc. and that's the important part.

You don't want to handle it on the main thread, because it may be busy executing
blocking commands (e.g. glFinish etc).

[David Yen, 2015/09/28 17:38:18]

Done.

     IPC_MESSAGE_HANDLER_DELAY_REPLY(GpuMsg_CreateJpegDecoder,
                                     OnCreateJpegDecoder)
     IPC_MESSAGE_UNHANDLED(handled = false)
@@ -1021,6 +1023,8 @@ void GpuChannel::OnCreateJpegDecoder(int32 route_id, IPC::Message* reply_msg) {
   jpeg_decoder_->AddClient(route_id, reply_msg);
 }
 
+void GpuChannel::OnNop() {}
+
 void GpuChannel::CacheShader(const std::string& key,
                              const std::string& shader) {
   gpu_channel_manager_->Send(