Remove swap ack throttling to regain performance regression on freon

cc/scheduler/scheduler_state_machine.cc

Index: cc/scheduler/scheduler_state_machine.cc
diff --git a/cc/scheduler/scheduler_state_machine.cc b/cc/scheduler/scheduler_state_machine.cc
index c512987bd23f501abdaf1fceab783945f8dc1ae4..dd98a9e987f26c3b194b5d31b1028bb6722af057 100644
--- a/cc/scheduler/scheduler_state_machine.cc
+++ b/cc/scheduler/scheduler_state_machine.cc
@@ -425,11 +425,13 @@ bool SchedulerStateMachine::ShouldSendBeginMainFrame() const {
 
   // SwapAck throttle the BeginMainFrames unless we just swapped.
   // TODO(brianderson): Remove this restriction to improve throughput.
+#if !defined(ARCH_CPU_ARM_FAMILY)

[danakj, 2015/04/06 16:32:51]

This should be a LayerTreeSetting->SchedulerSetting set in the embedder. Please
no platform #ifs inside cc/. Then we can not unit test the paths easily.

[brianderson, 2015/04/07 03:15:38]

Let's just enable this on all platforms and see if there's still any browser
test flake. I can help you debug if there is.

I'm pretty sure the flake is just because of poorly written tests, but  on the
off chance that it's not, I'd hate for only one class of devices to be broken.

Enabling it on all devices will also give us some real feedback from the perf
bots concerning tradeoffs. There's a theoretical concern that this will increase
CPU fighting with the main thread. If it does, we may want to hold off on this
and be smarter about when we throttle vs. when we don't.

   bool just_swapped_in_deadline =
       begin_impl_frame_state_ == BEGIN_IMPL_FRAME_STATE_INSIDE_DEADLINE &&
       perform_swap_funnel_;
   if (pending_swaps_ >= max_pending_swaps_ && !just_swapped_in_deadline)
     return false;
+#endif
 
   if (skip_begin_main_frame_to_reduce_latency_)
     return false;