Pass -Oz flag to Clang when targeting Android.

Pass -Oz flag to Clang when targeting Android.

The -Oz flag enables more aggressive size optimizations. Shrinks the
libchromeshell.so DSO by around 2.5MB.

BUG=481675
R=thakis@chromium.org

Committed: https://crrev.com/2433e6ab2206fad3bc5954fd61e9d4785ccab108
Cr-Commit-Position: refs/heads/master@{#327183}

[Nico, 2015-04-27 22:55:08]

lgtm

[pcc, 2015-04-27 22:59:41]

The CQ bit was checked by pcc@google.com

[commit-bot: I haz the power, 2015-04-27 23:00:28]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1103393002/20001

[jdduke (slow), 2015-04-27 23:10:38]

On 2015/04/27 23:00:28, I haz the power (commit-bot) wrote:
> CQ is trying da patch. Follow status at
>  https://chromium-cq-status.appspot.com/patch-status/1103393002/20001

What kind of perf impact does this have? While we'd like to optimize size,
ideally we'd do it in a way that minimizes any perf regressions. Have you run
any relevant telemetry perf tests?

[Nico, 2015-04-27 23:14:08]

On 2015/04/27 23:10:38, jdduke wrote:
> On 2015/04/27 23:00:28, I haz the power (commit-bot) wrote:
> > CQ is trying da patch. Follow status at
> >  https://chromium-cq-status.appspot.com/patch-status/1103393002/20001
> 
> What kind of perf impact does this have? While we'd like to optimize size,
> ideally we'd do it in a way that minimizes any perf regressions. Have you run
> any relevant telemetry perf tests?

We don't use clang for any production builds on Android, so this isn't a concern
at this point.

(If we switch from gcc to clang, we want to be competitive in both size and perf
of course.)

[jdduke (slow), 2015-04-27 23:17:46]

On 2015/04/27 23:14:08, Nico wrote:
> On 2015/04/27 23:10:38, jdduke wrote:
> > On 2015/04/27 23:00:28, I haz the power (commit-bot) wrote:
> > > CQ is trying da patch. Follow status at
> > >  https://chromium-cq-status.appspot.com/patch-status/1103393002/20001
> > 
> > What kind of perf impact does this have? While we'd like to optimize size,
> > ideally we'd do it in a way that minimizes any perf regressions. Have you
run
> > any relevant telemetry perf tests?
> 
> We don't use clang for any production builds on Android, so this isn't a
concern
> at this point.
> 
> (If we switch from gcc to clang, we want to be competitive in both size and
perf
> of course.)

I know, and the question still stands. If this regresses clang perf by, say, 5%
in some key metrics, it may not be worth it and we should consider alternative
approaches to achieve a competive clang APK size.

[Nico, 2015-04-27 23:45:25]

On 2015/04/27 23:17:46, jdduke wrote:
> On 2015/04/27 23:14:08, Nico wrote:
> > On 2015/04/27 23:10:38, jdduke wrote:
> > > On 2015/04/27 23:00:28, I haz the power (commit-bot) wrote:
> > > > CQ is trying da patch. Follow status at
> > > >  https://chromium-cq-status.appspot.com/patch-status/1103393002/20001
> > > 
> > > What kind of perf impact does this have? While we'd like to optimize size,
> > > ideally we'd do it in a way that minimizes any perf regressions. Have you
> run
> > > any relevant telemetry perf tests?
> > 
> > We don't use clang for any production builds on Android, so this isn't a
> concern
> > at this point.
> > 
> > (If we switch from gcc to clang, we want to be competitive in both size and
> perf
> > of course.)
> 
> I know, and the question still stands.

No, it doesn't :-) If we manage to hit gcc's size and perf we can switch. We can
then experiment after that, but before we need to focus on matching gcc.

> If this regresses clang perf by, say, 5%
> in some key metrics, it may not be worth it and we should consider alternative
> approaches to achieve a competive clang APK size.

[commit-bot: I haz the power, 2015-04-28 00:03:41]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-04-28 00:05:40]

Patchset 2 (id:??) landed as
https://crrev.com/2433e6ab2206fad3bc5954fd61e9d4785ccab108
Cr-Commit-Position: refs/heads/master@{#327183}

[pcc, 2015-04-28 18:09:03]

A revert of this CL (patchset #2 id:20001) has been created in
https://codereview.chromium.org/1109203002/ by pcc@google.com.

The reason for reverting is: Causes a build failure with
-Dcomponent=shared_library.

[pasko, 2015-04-30 15:48:48]

pasko@chromium.org changed reviewers:
+ pasko@chromium.org

[pasko, 2015-04-30 15:48:49]

https://codereview.chromium.org/1103393002/diff/20001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1103393002/diff/20001/build/common.gypi#newco...
build/common.gypi:3747: 'release_optimize%': 'z',
counter-intuitively there is another file to be tweaked:
build/android/increase_size_for_speed.gypi