Make Telemetry report VM peak metric on Android.

Make Telemetry report VM peak metric on Android.

BUG=311633
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=233861

[tonyg, 2013-10-30 14:40:00]

Drive by: But I'm wondering why we care about VM usage. Can you explain any
scenario where we get useful info from it?

[Philippe, 2013-10-30 14:44:07]

On 2013/10/30 14:40:00, tonyg wrote:
> Drive by: But I'm wondering why we care about VM usage. Can you explain any
> scenario where we get useful info from it?

We are going to purge the unpinned ashmem regions before we do any measurement
therefore we won't see our usage of unpinned ashmem. The VM RSS peak will help
us guarantee that we don't have a crazy use of unpinned ashmem. While using a
lot of unpinned ashmem is not as bad as using a lot of private dirty memory,
using it excessively can still cause memory pressure which we want to avoid.
Does that make any sense?

[Philippe, 2013-10-30 14:46:27]

On 2013/10/30 14:44:07, Philippe wrote:
> On 2013/10/30 14:40:00, tonyg wrote:
> > Drive by: But I'm wondering why we care about VM usage. Can you explain any
> > scenario where we get useful info from it?
> 
> We are going to purge the unpinned ashmem regions before we do any measurement
> therefore we won't see our usage of unpinned ashmem. The VM RSS peak will help
> us guarantee that we don't have a crazy use of unpinned ashmem. While using a
> lot of unpinned ashmem is not as bad as using a lot of private dirty memory,
> using it excessively can still cause memory pressure which we want to avoid.
> Does that make any sense?

Besides that VM peak has already been a problem for us in the past. We had our
renderers killed by the OOM killer due to an excessive peak.

[bulach, 2013-11-04 13:10:01]

lgtm if tony is also convinced (I am! :)

just a small suggestion below:

https://chromiumcodereview.appspot.com/52503002/diff/1/tools/telemetry/teleme...
File tools/telemetry/telemetry/core/platform/android_platform_backend.py
(right):

https://chromiumcodereview.appspot.com/52503002/diff/1/tools/telemetry/teleme...
tools/telemetry/telemetry/core/platform/android_platform_backend.py:108:
'VMPeak': memory_usage['VmHighWaterMark'] * 1024}
in /proc this is VmHWM, then VmHighWaterMark and in the result it becomes
VMPeak..
I'd prefer keeping VmHWM all across the board and avoid these indirections...
seeing the Pss is expanded in line 105, perhaps just keeping as VmHighWaterMark
here?

[Philippe, 2013-11-04 14:00:43]

Thanks Marcus! Yeah, let's wait for Tony.

https://chromiumcodereview.appspot.com/52503002/diff/1/tools/telemetry/teleme...
File tools/telemetry/telemetry/core/platform/android_platform_backend.py
(right):

https://chromiumcodereview.appspot.com/52503002/diff/1/tools/telemetry/teleme...
tools/telemetry/telemetry/core/platform/android_platform_backend.py:108:
'VMPeak': memory_usage['VmHighWaterMark'] * 1024}
On 2013/11/04 13:10:01, bulach wrote:
> in /proc this is VmHWM, then VmHighWaterMark and in the result it becomes
> VMPeak..
> I'd prefer keeping VmHWM all across the board and avoid these indirections...
> seeing the Pss is expanded in line 105, perhaps just keeping as
VmHighWaterMark
> here?

Done.

[Philippe, 2013-11-05 08:45:27]

Tony: Ping :)


On Mon, Nov 4, 2013 at 3:00 PM, <pliard@chromium.org> wrote:

> Thanks Marcus! Yeah, let's wait for Tony.
>
>
>
> https://chromiumcodereview.appspot.com/52503002/diff/1/
> tools/telemetry/telemetry/core/platform/android_platform_backend.py
> File tools/telemetry/telemetry/core/platform/android_platform_backend.py
> (right):
>
> https://chromiumcodereview.appspot.com/52503002/diff/1/
> tools/telemetry/telemetry/core/platform/android_
> platform_backend.py#newcode108
> tools/telemetry/telemetry/core/platform/android_platform_backend.py:108:
> 'VMPeak': memory_usage['VmHighWaterMark'] * 1024}
> On 2013/11/04 13:10:01, bulach wrote:
>
>> in /proc this is VmHWM, then VmHighWaterMark and in the result it
>>
> becomes
>
>> VMPeak..
>> I'd prefer keeping VmHWM all across the board and avoid these
>>
> indirections...
>
>> seeing the Pss is expanded in line 105, perhaps just keeping as
>>
> VmHighWaterMark
>
>> here?
>>
>
> Done.
>
> https://chromiumcodereview.appspot.com/52503002/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[Philippe, 2013-11-05 11:14:17]

CC'ing Ross who is also interested in peak usage :)

[Philippe, 2013-11-06 08:37:27]

Tony? :)


On Tue, Nov 5, 2013 at 12:14 PM, <pliard@chromium.org> wrote:

> CC'ing Ross who is also interested in peak usage :)
>
> https://chromiumcodereview.appspot.com/52503002/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[tonyg, 2013-11-06 16:26:47]

lgtm

Really sorry for the delay, I got kind of backed up :/

[Philippe, 2013-11-06 16:34:52]

Thanks Tony! No worries :)


On Wed, Nov 6, 2013 at 5:26 PM, <tonyg@chromium.org> wrote:

> lgtm
>
> Really sorry for the delay, I got kind of backed up :/
>
> https://chromiumcodereview.appspot.com/52503002/
>

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[commit-bot: I haz the power, 2013-11-06 17:22:40]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/pliard@chromium.org/52503002/200001

[commit-bot: I haz the power, 2013-11-06 21:30:58]

Retried try job too often on win_rel for step(s) browser_tests,
chrome_frame_net_tests, chrome_frame_tests, chrome_frame_unittests,
content_browsertests, mini_installer_test, nacl_integration,
sync_integration_tests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...

[commit-bot: I haz the power, 2013-11-08 09:46:06]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/pliard@chromium.org/52503002/200001

[commit-bot: I haz the power, 2013-11-08 11:19:20]

Change committed as 233861