Version 3.29.88.5 (merged r24219)

Version 3.29.88.5 (merged r24219)

Enable FDO for v8 in android build.

R=danno@chromium.org, torne@chromium.org
BUG=

Committed: https://code.google.com/p/v8/source/detail?r=24517

[Torne, 2014-10-09 15:41:00]

LGTM, Thanks Ulan!

[ulan, 2014-10-09 15:54:57]

Torne, we usually do not merge back anything except critical bug fixes that have
canary coverage. Does this have any coverage?

Could you please explain here why you need this in 3.29 and why it is safe? (to
make it easier for me to get L G T M from V8 team).

[Torne, 2014-10-09 16:38:21]

On 2014/10/09 15:54:57, ulan wrote:
> Torne, we usually do not merge back anything except critical bug fixes that
have
> canary coverage. Does this have any coverage?
> 
> Could you please explain here why you need this in 3.29 and why it is safe?
(to
> make it easier for me to get L G T M from V8 team).

The people working on doing FDO for the android platform are intending to start
collecting profiling data and feeding it back into the build on an android
branch that is tracking chromium M39, and V8 is the most critical part of
chromium on which to enable FDO data collection/optimisation. We'll end up
cherrypicking this downstream if it doesn't land in V8 here, but we prefer to
keep the android branches unforked from upstream if we can.

It should be safe since all it does is introduce the FDO flag into the generated
makefiles on android webview builds, and the flag on its own does nothing (since
the FDO data collection/optimisation mode still has to be enabled globally in
the build, which will only happen for selected release builds). It's ultimately
up to the FDO compiler guys to test the results there.

[ulan, 2014-10-09 16:46:03]

ulan@chromium.org changed reviewers:
+ danno@chromium.org

[ulan, 2014-10-09 16:46:46]

On 2014/10/09 16:38:21, Torne wrote:
> On 2014/10/09 15:54:57, ulan wrote:
> > Torne, we usually do not merge back anything except critical bug fixes that
> have
> > canary coverage. Does this have any coverage?
> > 
> > Could you please explain here why you need this in 3.29 and why it is safe?
> (to
> > make it easier for me to get L G T M from V8 team).
> 
> The people working on doing FDO for the android platform are intending to
start
> collecting profiling data and feeding it back into the build on an android
> branch that is tracking chromium M39, and V8 is the most critical part of
> chromium on which to enable FDO data collection/optimisation. We'll end up
> cherrypicking this downstream if it doesn't land in V8 here, but we prefer to
> keep the android branches unforked from upstream if we can.
> 
> It should be safe since all it does is introduce the FDO flag into the
generated
> makefiles on android webview builds, and the flag on its own does nothing
(since
> the FDO data collection/optimisation mode still has to be enabled globally in
> the build, which will only happen for selected release builds). It's
ultimately
> up to the FDO compiler guys to test the results there.

Thanks, Torne. Adding danno@ to reviewers.

[danno, 2014-10-10 09:42:25]

lgtm

[ulan, 2014-10-10 09:43:56]

Committed patchset #1 (id:1) manually as 24517.