Lazily compute SecurityOrigin::m_isPotentiallyTrustworthy for performance

• Component: WebCore Misc.
• Status: RESOLVED FIXED

[Chris Dumez]

We currently initialize m_isPotentiallyTrustworthy in the SecurityOrigin constructor. However, it is a bit expensive to compute and shows on profiles, and we often don't actually use this data member. As a result, this patch makes it so that SecurityOrigin::m_isPotentiallyTrustworthy gets computed lazily.

[Chris Dumez]

Created attachment 431240 [details]
Patch

[Alex Christensen]

Comment on attachment 431240 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=431240&action=review

> Source/WebCore/page/SecurityOrigin.cpp:496
> +    return m_isPotentiallyTrustworthy;

return *m_isPotentiallyTrustworthy;

> Source/WebCore/page/SecurityOrigin.h:246
> +    std::optional<bool> m_isPotentiallyTrustworthy;

This uses more memory.  To offset this, could you make some of the surrounding booleans bit fields?

[Chris Dumez]

(In reply to Alex Christensen from comment #2)
> Comment on attachment 431240 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=431240&action=review
> 
> > Source/WebCore/page/SecurityOrigin.cpp:496
> > +    return m_isPotentiallyTrustworthy;
> 
> return *m_isPotentiallyTrustworthy;

Indeed, oops.

> 
> > Source/WebCore/page/SecurityOrigin.h:246
> > +    std::optional<bool> m_isPotentiallyTrustworthy;
> 
> This uses more memory.  To offset this, could you make some of the
> surrounding booleans bit fields?

sizeof(SecurityOrigin) is 64 before and after my change so let's not?

[Chris Dumez]

Created attachment 431242 [details]
Patch

[Chris Dumez]

Created attachment 431243 [details]
Patch

[Chris Dumez]

Created attachment 431245 [details]
Patch

[EWS]

Committed r278806 (238763@main): <https://commits.webkit.org/238763@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 431245 [details].

[Radar WebKit Bug Importer]

<rdar://problem/79228666>