Add performance test of character_fallback with AAT fonts

Add performance test of character_fallback with AAT fonts

crbug.com/547912 figured out that font fallback runs different code
path for AAT fonts. This patch adds a performance test that
specifically runs the code path.

BUG=547912
Committed: https://crrev.com/4e4de201f79029003a5e743f0893ce2e1f832e14
Cr-Commit-Position: refs/heads/master@{#372636}

[kojii, 2016-01-12 07:33:54]

kojii@chromium.org changed reviewers:
+ drott@chromium.org

[kojii, 2016-01-12 07:33:55]

drott@, should we do this?

[drott, 2016-01-12 08:36:49]

Thanks for looking into adapting the test. I'd prefer to set up a separate
performance test for that, perhaps character_fallback_aat.html, and ideally
restrict the tested characters to those that are missing from the first 4 fonts
(e.g. CJK, or using other unicode blocks that are not in the mentioned fonts
either). This would make the test a bit sharper and focused on the AAT issue. It
should be easy to adapt by specifiying a new .js file with a different array of
tested chars.

[kojii, 2016-01-19 02:13:32]

Description was changed from

==========
perf-mac-aat-fallback

BUG=547912
==========

to

==========
perf-mac-aat-fallback

BUG=547912
==========

[kojii, 2016-01-19 06:36:34]

Description was changed from

==========
perf-mac-aat-fallback

BUG=547912
==========

to

==========
Add performance test of character_fallback with AAT fonts

crbug.com/547912 figured out that font fallback runs different code
path for AAT fonts. This patch adds a performance test that
specifically runs the code path.

BUG=547912
==========

[kojii, 2016-01-19 06:37:19]

PTAL.

Didn't really looked into all CMAPs but limited to CJK and PUA, which should do
the most of what we need.

[drott, 2016-01-19 08:30:20]

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
File third_party/WebKit/PerformanceTests/Layout/character_fallback_aat.html
(right):

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
third_party/WebKit/PerformanceTests/Layout/character_fallback_aat.html:10:
font-family: Palatino, Georgia, Times, 'Times New Roman';
Palatino and Times are AAT, Georgia and Times New Roman are not, and Hiragino
Mincho Pro has glyphs.

So, to emphasize the AAT fallback, and arrive at a font that has coverage and
not go through to system font fallback, I think the font stack here should be:
"Palatino, Times, Hiragino Mincho Pro" 

(https://code.google.com/p/chromium/issues/detail?id=572104#c14)

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
File
third_party/WebKit/PerformanceTests/Layout/resources/character_fallback_chars_aat.js
(right):

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
third_party/WebKit/PerformanceTests/Layout/resources/character_fallback_chars_aat.js:1:
// This list of supported unicode characters was extracted from the Linux
This comment can be removed, only relevant for the previous generated character
file for which I previously did a CMAP coverage analysis on Linux.

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
third_party/WebKit/PerformanceTests/Layout/resources/character_fallback_chars_aat.js:2082:
fallbackChars = fallbackChars.concat(charsSupportedByNFonts[j]);
This assembly of fallbackChars from the charsSupportedByNFonts is not really
needed here since we don't have the "charsSupportedByNFonts" histogram
information on Mac without a detailed CMAP analysis. So just putting CJK unicode
ranges programmatically into fallbackChars is sufficient here, based on the
assumption that Palatino and Times do not have coverage for them.

[kojii, 2016-01-25 05:25:16]

PTAL.

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
File third_party/WebKit/PerformanceTests/Layout/character_fallback_aat.html
(right):

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
third_party/WebKit/PerformanceTests/Layout/character_fallback_aat.html:10:
font-family: Palatino, Georgia, Times, 'Times New Roman';
On 2016/01/19 08:30:20, drott wrote:
> Palatino and Times are AAT, Georgia and Times New Roman are not, and Hiragino
> Mincho Pro has glyphs.
> 
> So, to emphasize the AAT fallback, and arrive at a font that has coverage and
> not go through to system font fallback, I think the font stack here should be:
> "Palatino, Times, Hiragino Mincho Pro" 
> 
> (https://code.google.com/p/chromium/issues/detail?id=572104#c14)
> 

Done.

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
File
third_party/WebKit/PerformanceTests/Layout/resources/character_fallback_chars_aat.js
(right):

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
third_party/WebKit/PerformanceTests/Layout/resources/character_fallback_chars_aat.js:1:
// This list of supported unicode characters was extracted from the Linux
On 2016/01/19 08:30:20, drott wrote:
> This comment can be removed, only relevant for the previous generated
character
> file for which I previously did a CMAP coverage analysis on Linux.

Done.

https://codereview.chromium.org/1578913003/diff/20001/third_party/WebKit/Perf...
third_party/WebKit/PerformanceTests/Layout/resources/character_fallback_chars_aat.js:2082:
fallbackChars = fallbackChars.concat(charsSupportedByNFonts[j]);
On 2016/01/19 08:30:20, drott wrote:
> This assembly of fallbackChars from the charsSupportedByNFonts is not really
> needed here since we don't have the "charsSupportedByNFonts" histogram
> information on Mac without a detailed CMAP analysis. So just putting CJK
unicode
> ranges programmatically into fallbackChars is sufficient here, based on the
> assumption that Palatino and Times do not have coverage for them.

Done.

[drott, 2016-02-01 07:51:10]

LGTM

[kojii, 2016-02-01 10:07:02]

The CQ bit was checked by kojii@chromium.org

[commit-bot: I haz the power, 2016-02-01 10:07:52]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1578913003/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1578913003/40001

[commit-bot: I haz the power, 2016-02-01 10:57:33]

Description was changed from

==========
Add performance test of character_fallback with AAT fonts

crbug.com/547912 figured out that font fallback runs different code
path for AAT fonts. This patch adds a performance test that
specifically runs the code path.

BUG=547912
==========

to

==========
Add performance test of character_fallback with AAT fonts

crbug.com/547912 figured out that font fallback runs different code
path for AAT fonts. This patch adds a performance test that
specifically runs the code path.

BUG=547912
==========

[commit-bot: I haz the power, 2016-02-01 10:57:34]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-02-01 10:59:01]

Description was changed from

==========
Add performance test of character_fallback with AAT fonts

crbug.com/547912 figured out that font fallback runs different code
path for AAT fonts. This patch adds a performance test that
specifically runs the code path.

BUG=547912
==========

to

==========
Add performance test of character_fallback with AAT fonts

crbug.com/547912 figured out that font fallback runs different code
path for AAT fonts. This patch adds a performance test that
specifically runs the code path.

BUG=547912

Committed: https://crrev.com/4e4de201f79029003a5e743f0893ce2e1f832e14
Cr-Commit-Position: refs/heads/master@{#372636}
==========

[commit-bot: I haz the power, 2016-02-01 10:59:01]

Patchset 3 (id:??) landed as
https://crrev.com/4e4de201f79029003a5e743f0893ce2e1f832e14
Cr-Commit-Position: refs/heads/master@{#372636}