Convert memory_test to use the common stats-reporting methods in UITest....

Convert memory_test to use the common stats-reporting methods in UITest.
This adds IO stats to memory_test, and renames most of the memory_test
traces to match the names used by page_cycler so that (after the historical
memory_ data is converted) we can do apples-to-apples comparisons
of stats across all the tests that use these methods.
BUG=none
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=32632

[sgk, 2009-11-20 00:09:06]

If this is all right, I'd convert memory_test stats by hand to preserve history.
 I'm waiting on trybot success, and still need to verify comparability of stats
and trace names.

[Mike Belshe, 2009-11-20 10:27:13]

http://codereview.chromium.org/402104/diff/3001/4002
File chrome/test/ui/ui_test.cc (right):

http://codereview.chromium.org/402104/diff/3001/4002#newcode1337
Line 1337: browser_peak_virtual_size += peak_virtual_size;
do you really want to *add* the peak_virtual_size?  Or just set it?

http://codereview.chromium.org/402104/diff/3001/4002#newcode1340
Line 1340: renderer_peak_virtual_size += peak_virtual_size;
ditto - check "if (peak > current_peak) current_peak = peak"?

[sgk, 2009-11-20 17:25:36]

http://codereview.chromium.org/402104/diff/3001/4002
File chrome/test/ui/ui_test.cc (right):

http://codereview.chromium.org/402104/diff/3001/4002#newcode1337
Line 1337: browser_peak_virtual_size += peak_virtual_size;
On 2009/11/20 10:27:13, Mike Belshe wrote:
> do you really want to *add* the peak_virtual_size?  Or just set it?

You're right, should be = because we're just recording the browser pid's peak. 
Done.

http://codereview.chromium.org/402104/diff/3001/4002#newcode1340
Line 1340: renderer_peak_virtual_size += peak_virtual_size;
On 2009/11/20 10:27:13, Mike Belshe wrote:
> ditto - check "if (peak > current_peak) current_peak = peak"?

If we're running a test with multiple renderers, recording only the peak of the
largest renderer would lose information.  If we had 20 renderers and the largest
had the same peak as last run, but the smaller 19 all had large increases in
their peak size, I think we'd want to know that.

The maximum peak size of any single renderer would also be useful to know,
though.  I'll add a separate stat for that.

[Mike Belshe, 2009-11-20 18:20:37]

lgtm