RESOLVED FIXED292696
[Yarr] Improve efficiency of non-BMP character classes with a fixed size greater than 1
https://bugs.webkit.org/show_bug.cgi?id=292696
Summary [Yarr] Improve efficiency of non-BMP character classes with a fixed size grea...
Michael Saboff
Reported 2025-05-07 17:35:02 PDT
When reading surrogate pairs, we have an optimization to advance 2 UTF-16 characters when the first code point we read is a non-BMP code point. That optimization involves checking the first code point loaded in the top level disjunction. That optimization uses a small set of instructions to check if we read the first code point and its size. For a character class consisting of only non-BMP characters, we don't need to perform the 'is this the first code point read and what its size" check when reading code points after the first code point.
Attachments
Radar WebKit Bug Importer
Comment 1 2025-05-07 17:35:53 PDT
Michael Saboff
Comment 2 2025-05-07 18:04:47 PDT
Michael Saboff
Comment 3 2025-05-12 11:14:39 PDT
EWS
Comment 4 2025-05-13 09:39:38 PDT
Committed 294852@main (608810ff3697): <https://commits.webkit.org/294852@main> Reviewed commits have been landed. Closing PR #45260 and removing active labels.
Note You need to log in before you can comment on or make changes to this bug.