Remove 1 ms rounding up of task delays on Windows

Remove 1 ms rounding up of task delays on Windows

Prior to a couple of attempts at fixing bug 487724 the message pump
would sometimes end up spinning for up to a ms when waiting for a
message's time to arrive. This was unintentional and non-obvious.

The eventual fix was to round up the delay time to the next ms.
Unfortunately this caused a regression in smoothness.top_25_smooth
which has not been resolved (investigation was delayed due to some
confusion about the results). Therefore this change reverts both
fix attempts in order to leave time for a proper investigation and
any necessary fixes.

It is worth mentioning that the delay introduced by rounding up to 1
ms is actually worse than expected on some operating systems. On Windows
7 a call to Sleep(1) will, if the timer frequency is raised, return
within a ms. On Windows 8.1 it will often take two ms to return, which
increases the delay. If the timer frequency is at its default interval
of 15.625 ms then the delay can be longer.

BUG=487724, 497536
Committed: https://crrev.com/2b1bf190a079709a1a6a769bf86e782423d23121
Cr-Commit-Position: refs/heads/master@{#344688}

[brucedawson, 2015-08-20 22:10:28]

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

[brucedawson, 2015-08-20 22:13:40]

After discussion with rschoen@, amineer@, and tinazh@ I am reverting the last
couple of attempts at avoiding some CPU/power waste in Chrome's message loop.
The plan is to merge this to M45, and then have time to understand the one
regression which this caused.

This will return the code to its state prior to May 14th.

[rvargas (doing something else), 2015-08-20 22:31:15]

lgtm

[brucedawson, 2015-08-20 22:44:38]

Moving this from the description:

It was confirmed that this change completely undoes the previous two
with these git commands:

>git log -4 --oneline base\message_loop\message_pump_default.cc
81a76bc Remove 1 ms rounding up of task delays
932c245 Set minimum Windows wait time to 1 ms in MessagePumpDefault::Run
990d0c7 Treat a delay of under 1 millisecond in MessagePumpDefault
54aa4f1 Made MessagePump a non-thread safe class.

git diff 81a76bc..54aa4f1 -- base\message_loop\message_pump_default.cc

[brucedawson, 2015-08-20 22:44:44]

The CQ bit was checked by brucedawson@chromium.org

[commit-bot: I haz the power, 2015-08-20 22:47:56]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1305873002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1305873002/1

[commit-bot: I haz the power, 2015-08-21 05:42:38]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-08-21 05:43:27]

Patchset 1 (id:??) landed as
https://crrev.com/2b1bf190a079709a1a6a769bf86e782423d23121
Cr-Commit-Position: refs/heads/master@{#344688}

[pfeldman, 2015-08-21 17:09:23]

A revert of this CL (patchset #1 id:1) has been created in
https://codereview.chromium.org/1304293002/ by pfeldman@chromium.org.

The reason for reverting is: Figuring out if this could cause Win10 failures:
http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win10/builds/71.

[brucedawson, 2015-08-21 18:48:25]

Did revert this make any difference? It seems unlikely that my change could have
caused these failures. Let me know as I'd like to re-land the change today if
possible.