Use the ASCII-specific identifier-part predicate in TokenStreamSpecific::identifierName for the hot ASCII identifier-continuation scanning path.
Categories
(Core :: Performance: Automated Findings, defect)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox160 | --- | fixed |
People
(Reporter: perf-ai-bot, Assigned: iain)
References
(Blocks 1 open bug)
Details
(Whiteboard: [foofrix:12e8729b-dd59-4b0f-829e-8607d0bfcf67])
Attachments
(1 external file, 1 obsolete file)
Use the ASCII-specific identifier-part predicate in TokenStreamSpecific::identifierName for the hot ASCII identifier-continuation scanning path.
Problem
TokenStreamSpecific::identifierName calls the generic char16_t IsIdentifierPart helper for each ASCII identifier continuation code unit after the first character, adding an avoidable ASCII-range branch in the hot TipTap module parsing path.
Performance
- Benchmark: speedometer3
- Component: JS
- Local Improvement: 1.94%
Validation
| Reporter | ||
Updated•1 month ago
|
| Reporter | ||
Comment 1•1 month ago
|
||
Please move this bug to the correct component.
Once the Phabricator patch is ready for review, please commandeer it.
| Reporter | ||
Comment 2•1 month ago
|
||
Use the ASCII-specific identifier-part predicate in TokenStreamSpecific::identifierName for the hot ASCII identifier-continuation scanning path.
| Assignee | ||
Updated•1 month ago
|
| Assignee | ||
Comment 3•1 month ago
|
||
Foofrix wrote this patch to avoid the if (ch < 128) branch in IsIdentifierPart. It seems like LLVM should be able to do this itself, but this does make my local build 1% faster in multi-inspector-code-load (which does a lot of parsing). It's plausible that PGO covers this, but this patch can't hurt.
Updated•1 month ago
|
Comment 5•1 month ago
|
||
| bugherder | ||
Updated•22 days ago
|
Description
•