cros: Add UMA stat for graphics driver memory on tab discard

cros: Add UMA stats for graphics driver memory

* Log graphics driver memory in the regular UMA memory stats
* Log also on tab discard
* Fix range of Tab.Discard.MemAnonymousMB to give more precision in the > 2 GB range
* Incorporate both Mali and GEM memory for ARM stats

BUG=164761, 164106
TEST=manual

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=172405

[James Cook, 2012-12-07 00:54:32]

Dave, PTAL.

[James Cook, 2012-12-07 18:05:29]

Dave, updated, PTAL.

[DaveMoore, 2012-12-07 18:52:48]

https://codereview.chromium.org/11475016/diff/5/chrome/browser/chromeos/memor...
File chrome/browser/chromeos/memory/oom_priority_manager.cc (right):

https://codereview.chromium.org/11475016/diff/5/chrome/browser/chromeos/memor...
chrome/browser/chromeos/memory/oom_priority_manager.cc:118: memory.gem_size /
1024 / 1024);
Use ui::FormatBytes() instead. You can find it in
ui/base/text/bytes_formatting.h

https://codereview.chromium.org/11475016/diff/5/chrome/browser/chromeos/memor...
chrome/browser/chromeos/memory/oom_priority_manager.cc:304: if (memory.gem_size
!= -1) {
If you do this on Intel you will double count memory. On ARM that's not the
case. I don't know why, but the drivers appear to allocate the memory
differently

https://codereview.chromium.org/11475016/diff/5/chrome/browser/ui/webui/about...
File chrome/browser/ui/webui/about_ui.cc (right):

https://codereview.chromium.org/11475016/diff/5/chrome/browser/ui/webui/about...
chrome/browser/ui/webui/about_ui.cc:384: output.append(AddStringRow(
Is this automatically added to a total? If so there will be a problem on Intel.

[James Cook, 2012-12-07 19:41:22]

Updated, PTAL.

https://codereview.chromium.org/11475016/diff/5/chrome/browser/chromeos/memor...
File chrome/browser/chromeos/memory/oom_priority_manager.cc (right):

https://codereview.chromium.org/11475016/diff/5/chrome/browser/chromeos/memor...
chrome/browser/chromeos/memory/oom_priority_manager.cc:118: memory.gem_size /
1024 / 1024);
On 2012/12/07 18:52:48, DaveMoore wrote:
> Use ui::FormatBytes() instead. You can find it in
> ui/base/text/bytes_formatting.h

Done.

https://codereview.chromium.org/11475016/diff/5/chrome/browser/chromeos/memor...
chrome/browser/chromeos/memory/oom_priority_manager.cc:304: if (memory.gem_size
!= -1) {
On 2012/12/07 18:52:48, DaveMoore wrote:
> If you do this on Intel you will double count memory. On ARM that's not the
> case. I don't know why, but the drivers appear to allocate the memory
> differently

Replaced with a MemAllocatedMB stat that tracks both anonymous pages from
malloc() and non-anonymous graphics driver pages on ARM.

https://codereview.chromium.org/11475016/diff/5/chrome/browser/ui/webui/about...
File chrome/browser/ui/webui/about_ui.cc (right):

https://codereview.chromium.org/11475016/diff/5/chrome/browser/ui/webui/about...
chrome/browser/ui/webui/about_ui.cc:384: output.append(AddStringRow(
On 2012/12/07 18:52:48, DaveMoore wrote:
> Is this automatically added to a total? If so there will be a problem on
Intel.

It's not adding to any other totals in the UI. There will be double-counting on
Intel in the active_anon and inactive_anon above (not sure if in one or both),
but since this is a developer page trying to approximate a dump of /proc/meminfo
I'd like to leave it as-is.

[James Cook, 2012-12-07 19:49:34]

Changed my mind, I'm adding allocated to about:discards, patch to follow.

[James Cook, 2012-12-07 19:51:29]

Updated, PTAL.

[James Cook, 2012-12-10 23:01:04]

Ping Dave?

[DaveMoore, 2012-12-11 05:20:40]

lgtm

[James Cook, 2012-12-11 18:17:13]

brettw, can I get OWNERS approval for base/process_util_linux.cc?  We've had
some trouble with graphics memory consumption on Chrome OS and this extends
Dave's memory tracking code to give us better visibility on ARM/Exynos (the new
Chromebook).

[brettw, 2012-12-11 19:16:47]

base owners lgtm, hopefully dave understands this because I have no idea what's
going on.

[commit-bot: I haz the power, 2012-12-11 20:04:36]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jamescook@chromium.org/11475016/15001

[commit-bot: I haz the power, 2012-12-11 21:24:41]

Change committed as 172405