Bug 92677
| Summary: | [meta] ascii string marking | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Dan Carney <dcarney> |
| Component: | WebCore Misc. | Assignee: | Nobody <webkit-unassigned> |
| Status: | UNCONFIRMED | ||
| Severity: | Normal | CC: | jochen, msaboff, peter |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Bug Depends on: | 92982 | ||
| Bug Blocks: | |||
Dan Carney
This metabug is to track work for implementing detection and marking of ascii strings during latin1 string creation. Please add specific implementation bugs that block this.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Michael Saboff
What problem is this work meant to fix?
jochen
(In reply to comment #1)
> What problem is this work meant to fix?
I'm summary (there was a mail to webkit-dev with a longer version):
V8 has several optimizations for ASCII strings, i.e. you can export ascii to utf8 by memcpy (but not latin1), regexp matching is faster, toUpper(ascii) maps to ascii (but toUpper(latin1) does not map to latin1).
Detecting whether a string is ascii at creation time comes at no additional costs compared to detection latin1 which we already do. If we track ascii-ness of strings, we can use those optimizations.
JSC could benefit from this information as well.