diff options
| author | sky@chromium.org <sky@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2009-09-14 17:19:29 +0000 |
|---|---|---|
| committer | sky@chromium.org <sky@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2009-09-14 17:19:29 +0000 |
| commit | f5c655de4121427b4d9d9dc218efd29afa8bba8c (patch) | |
| tree | d731b3ec3b974c96d21c54610b529c474a0fd7b2 /net | |
| parent | 2e360844a496013fc57c527fd5d6c4da165b09f7 (diff) | |
| download | chromium_src-f5c655de4121427b4d9d9dc218efd29afa8bba8c.zip chromium_src-f5c655de4121427b4d9d9dc218efd29afa8bba8c.tar.gz chromium_src-f5c655de4121427b4d9d9dc218efd29afa8bba8c.tar.bz2 | |
Attempt 2 at converting font from Skia to Pango. The slow down appears to be entirely attributed to asking for the metrics. To get back the performance I did the following:
. Only ask for the metrics when needed. We appear to create 6 or so fonts very early on without needing the metrics. This ensures we only take the metrics load hit when needed.
. Keep a cache of the metrics around. This ensures that we only load the metrics once.
I didn't dig through the Pango source to see why Pango is slow at getting metrics. I suspect it's the first person to ask for metrics incurs a penalty as Pango accesses the disk. The code we have now is the same as that as Gtk.
Review URL: http://codereview.chromium.org/199101
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@26116 0039d316-1c4b-4281-b951-d872f2087c98
Diffstat (limited to 'net')
0 files changed, 0 insertions, 0 deletions