Removed mtune for Android build.

Removed mtune for Android build.

mtune was set to cortex-a8, which is a Nexus S-generation CPU.
Removing it allows for small speed improvements (~1%) for
warm page load times on top10 mobile sites on more recent devices.
Data: https://docs.google.com/spreadsheet/ccc?key=0Ao-xIDuezqMNdDc5OWpxbkk4NFJ0SF9Iemt0Tlg3RFE&usp=sharing
And it has virtually no effect on binary size.
BUG=334271
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=247882

[Fabrice de Gans (Google), 2014-01-23 13:33:45]

brettw@chromium.org: Please review changes in 
build/config/arm.gni
ben@chromium.org: Please review changes in 
build/common.gypi

[pasko, 2014-01-24 14:21:05]

I'd like to add some more context.

The cortex-a8 flag was chosen for old ChromeOS ARM hardware in 2009 with this
change:
https://codereview.chromium.org/164208

Since piman@ was the reviewer for that change, it'd be best if he were the main
reviewer for this CL.

The improvements fdegans@ mentines is on Nexus 4/5. Given the amount of noise in
the measurements the 1% difference is on the boundary of insignificance. We will
keep an eye on performance bots/tests on other platforms. Heads up to perf
sheriffs before will be sent before landing this since we did not experiment
with all tests/platforms. Is there any particular perf testing reasonable to
perform now?

[pasko, 2014-01-24 14:23:10]

+ a few more ppl who might have an opinion

[piman, 2014-01-24 20:23:05]

Maybe Bhaskar has an opinon for Chrome OS. Ideally we'd specify the -mtune on a
per-board basis, in the Chrome OS tree. I don't know if we do that today.
But we're not shipping cortex-a8 anyway, so I'm not sure it's better than
anything else.

[brettw, 2014-01-25 12:47:02]

.gni file lgtm, although I have no opinion on whether this is a good change for
the build. I defer to piman.

[bjanakiraman1, 2014-01-26 00:51:27]

On 2014/01/24 20:23:05, piman wrote:
> Maybe Bhaskar has an opinon for Chrome OS. Ideally we'd specify the -mtune on
a
> per-board basis, in the Chrome OS tree. I don't know if we do that today.
> But we're not shipping cortex-a8 anyway, so I'm not sure it's better than
> anything else.


We use '-march=armv7-a -mtune=cortex-a15 -mfpu=neon -mfloat-abi=hard' by default
for all ChromeOS ARM boards shipping today (including big-little cores),  but as
piman said, we can customize our build for each board. 

I'm not sure what the question is. Removing -mtune= leaves it at the default
setting, which is whatever the toolchain/NDK is set up to do. I'd suggest trying
-mtune=cortex-A15 (but you'd need to evaluate the performance on A8 to make sure
its not too bad..)

[piman, 2014-01-27 18:52:37]

Given Bhaskar's response, this would have no effect on Chrome OS, so this LGTM.

[commit-bot: I haz the power, 2014-01-29 18:03:46]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/fdegans@chromium.org/145663002/1

[commit-bot: I haz the power, 2014-01-29 18:56:29]

Retried try job too often on mac_rel for step(s) remoting_unittests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

[commit-bot: I haz the power, 2014-01-29 19:09:00]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/fdegans@chromium.org/145663002/1

[commit-bot: I haz the power, 2014-01-30 07:53:46]

Change committed as 247882