Bug 161809 – [Cocoa] Reduce uses of CGFonts in favor of CTFonts

• Component: New Bugs
• Status: RESOLVED FIXED

[Myles C. Maxfield]

[Cocoa] Improve performance of populating glyph caches in the fast text codepath

[Myles C. Maxfield]

Created attachment 288424 [details]
WIP

[Myles C. Maxfield]

Created attachment 288430 [details]
WIP

[WebKit Commit Bot]

Attachment 288430 [details] did not pass style-queue:


ERROR: Source/WebCore/platform/graphics/mac/GlyphPageMac.cpp:80:  One line control clauses should not use braces.  [whitespace/braces] [4]
ERROR: Source/WebCore/ChangeLog:8:  You should remove the 'No new tests' and either add and list tests, or explain why no new tests were possible.  [changelog/nonewtests] [5]
Total errors found: 2 in 5 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Myles C. Maxfield]

Created attachment 288445 [details]
WIP

[WebKit Commit Bot]

Attachment 288445 [details] did not pass style-queue:


ERROR: Source/WebCore/platform/graphics/mac/GlyphPageMac.cpp:109:  One line control clauses should not use braces.  [whitespace/braces] [4]
ERROR: Source/WebCore/ChangeLog:8:  You should remove the 'No new tests' and either add and list tests, or explain why no new tests were possible.  [changelog/nonewtests] [5]
Total errors found: 2 in 7 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Myles C. Maxfield]

Created attachment 288446 [details]
WIP

[WebKit Commit Bot]

Attachment 288446 [details] did not pass style-queue:


ERROR: Source/WebCore/platform/graphics/mac/GlyphPageMac.cpp:110:  One line control clauses should not use braces.  [whitespace/braces] [4]
ERROR: Source/WebCore/ChangeLog:8:  You should remove the 'No new tests' and either add and list tests, or explain why no new tests were possible.  [changelog/nonewtests] [5]
Total errors found: 2 in 7 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Myles C. Maxfield]

Created attachment 288470 [details]
WIP

[Myles C. Maxfield]

Created attachment 288489 [details]
WIP

[Myles C. Maxfield]

Created attachment 288496 [details]
ASSERTs testing

[Myles C. Maxfield]

Created attachment 288497 [details]
Measuring perf

[Myles C. Maxfield]

Created attachment 288555 [details]
No regression

[Myles C. Maxfield]

Created attachment 288577 [details]
Patch

[Daniel Bates]

Comment on attachment 288577 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=288577&action=review

> Source/WTF/wtf/unicode/CharacterNames.h:73
> +const UChar narrowNoBreakSpace = 0x202F;

Please add this definitions such that it appears in this list of definitions in sorted order according to the UNIX sort command.

> Source/WTF/wtf/unicode/CharacterNames.h:136
> +using WTF::Unicode::narrowNoBreakSpace;

Ditto.

> Source/WebCore/platform/graphics/GlyphPage.h:45
> +    GlyphData(Glyph glyph = 0, const Font* font = 0)

We should take this opportunity to change 0 to nullptr for the default value of the second argument.

> Source/WebCore/platform/graphics/mac/GlyphPageMac.cpp:43
> +static bool shouldFillWithVerticalGlyphs(const UChar* buffer, unsigned bufferLength, const Font& font)

We may want to consider having this function take a WTF::StringView (or maybe introduce a WTF::ArrayView) instead of passing a buffer and length. Among other benefits, using a WTF::StringView helps ensure that we always have the correct buffer length for the buffer. I know that currently the caller GlyphPage::fill() is passed a raw pointer to a buffer and a length. We may want to look to change these signatures as well.

> Source/WebCore/platform/graphics/mac/GlyphPageMac.cpp:49
> +    if (!font.hasVerticalGlyphs())
> +        return false;
> +    for (unsigned i = 0; i < bufferLength; ++i) {
> +        if (!FontCascade::isCJKIdeograph(buffer[i]))
> +            return true;

I am assuming that we no longer need to consider whether the font is a system font or the font is for text combine as part of knowing whether a font has vertical glyphs.

[Myles C. Maxfield]

Comment on attachment 288577 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=288577&action=review

>> Source/WebCore/platform/graphics/mac/GlyphPageMac.cpp:43
>> +static bool shouldFillWithVerticalGlyphs(const UChar* buffer, unsigned bufferLength, const Font& font)
> 
> We may want to consider having this function take a WTF::StringView (or maybe introduce a WTF::ArrayView) instead of passing a buffer and length. Among other benefits, using a WTF::StringView helps ensure that we always have the correct buffer length for the buffer. I know that currently the caller GlyphPage::fill() is passed a raw pointer to a buffer and a length. We may want to look to change these signatures as well.

StringView is not a good candidate because it can hold both 8-bit characters and 16-bit characters, but this function only accepts 16-bit characters. Something like an ArrayView would be a better fit, but I think implementing that is out of scope for this patch.

>> Source/WebCore/platform/graphics/mac/GlyphPageMac.cpp:49
>> +            return true;
> 
> I am assuming that we no longer need to consider whether the font is a system font or the font is for text combine as part of knowing whether a font has vertical glyphs.

In the (deleted) code below, we only call CTFontGetVerticalGlyphsForCharacters() if the font is not a system font and the font is not for text combine.

[Myles C. Maxfield]

Committed r205826: <http://trac.webkit.org/changeset/205826>