Ensure that the wake up timer set by the UI worker thread in the UI message pump on Windows is set …

Ensure that the wake up timer set by the UI worker thread in the UI message pump on Windows is set once in each iteration.

Currently the wake up timer is set by the UI worker thread continuosly which causes the thread to wake up needlessly thus triggering CPU usage.

Fix is to set the timer from the worker thread once during each iteration.

BUG=500749

[ananta, 2015-06-18 13:51:43]

ananta@chromium.org changed reviewers:
+ scottmg@chromium.org

[scottmg, 2015-06-18 18:00:02]

scottmg@chromium.org changed reviewers:
+ brucedawson@chromium.org

[brucedawson, 2015-06-19 21:09:02]

See my comments on the code for one possible change to a comment, but not
required.

I tested with a local build and the steady wakeups are gone. I will check again
with a canary build after this is landed.

lgtm

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
File base/message_loop/message_pump_win.cc (right):

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
base/message_loop/message_pump_win.cc:444: // timeBeginPeriod API being called.
"is dependent on the current global timer precision and therefore depends on
whether Chrome or any other process has raised the timer frequency with
timeBeginPeriod."

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
File base/message_loop/message_pump_win.h (right):

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
base/message_loop/message_pump_win.h:178: long force_fallback_timer_for_tasks_;
I believe we are now allowed to use in-place initialization. If so this could be
tagged with "= 0;" instead of having to modify the constructor. I'm a huge fan.

FWIW.

[ananta, 2015-06-19 22:19:10]

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
File base/message_loop/message_pump_win.cc (right):

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
base/message_loop/message_pump_win.cc:444: // timeBeginPeriod API being called.
On 2015/06/19 21:09:01, brucedawson wrote:
> "is dependent on the current global timer precision and therefore depends on
> whether Chrome or any other process has raised the timer frequency with
> timeBeginPeriod."

Done.

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
File base/message_loop/message_pump_win.h (right):

https://codereview.chromium.org/1189133003/diff/1/base/message_loop/message_p...
base/message_loop/message_pump_win.h:178: long force_fallback_timer_for_tasks_;
On 2015/06/19 21:09:01, brucedawson wrote:
> I believe we are now allowed to use in-place initialization. If so this could
be
> tagged with "= 0;" instead of having to modify the constructor. I'm a huge
fan.
> 
> FWIW.

Leaving this as is for consistency for now.