Bug 226000 - AbstractSlotVisitor::containsOpaqueRoot() should only declare didFindOpaqueRoot if the root is actually found.
Summary: AbstractSlotVisitor::containsOpaqueRoot() should only declare didFindOpaqueRo...
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Mark Lam
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2021-05-19 19:24 PDT by Mark Lam
Modified: 2021-05-19 20:34 PDT (History)
7 users (show)

See Also:

Attachments
proposed patch. (1.64 KB, patch)
2021-05-19 19:29 PDT, Mark Lam 		no flags Details | Formatted Diff | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Mark Lam 2021-05-19 19:24:31 PDT 
It was erroneously calling didFindOpaqueRoot() all the time even when the root is not found.

rdar://78208014
Comment 1 Mark Lam 2021-05-19 19:29:55 PDT 
Created attachment 429126 [details]
proposed patch.
Comment 2 Robin Morisset 2021-05-19 19:35:25 PDT 
Comment on attachment 429126 [details]
proposed patch.

View in context: https://bugs.webkit.org/attachment.cgi?id=429126&action=review

> Source/JavaScriptCore/heap/AbstractSlotVisitorInlines.h:136
> +    if (UNLIKELY(found && m_needsExtraOpaqueRootHandling)) {

Is there a reason to check found first and m_needsExtraOpaqueRootHandling second?

If anything, I would expect the entire check of m_opaqueRoots.contains(ptr) to be only done after checking m_needsExtraOpaqueRootHandling, especially if it is UNLIKELY.
Comment 3 Mark Lam 2021-05-19 19:37:52 PDT 
(In reply to Robin Morisset from comment #2)
> Comment on attachment 429126 [details]
> proposed patch.
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=429126&action=review
> 
> > Source/JavaScriptCore/heap/AbstractSlotVisitorInlines.h:136
> > +    if (UNLIKELY(found && m_needsExtraOpaqueRootHandling)) {
> 
> Is there a reason to check found first and m_needsExtraOpaqueRootHandling
> second?
> 
> If anything, I would expect the entire check of m_opaqueRoots.contains(ptr)
> to be only done after checking m_needsExtraOpaqueRootHandling, especially if
> it is UNLIKELY.

I check found first because found is already in memory.  If not found, we save on triggering a load of m_needsExtraOpaqueRootHandling.
Comment 4 Robin Morisset 2021-05-19 19:43:14 PDT 
Comment on attachment 429126 [details]
proposed patch.

View in context: https://bugs.webkit.org/attachment.cgi?id=429126&action=review

r=me

>>> Source/JavaScriptCore/heap/AbstractSlotVisitorInlines.h:136
>>> +    if (UNLIKELY(found && m_needsExtraOpaqueRootHandling)) {
>> 
>> Is there a reason to check found first and m_needsExtraOpaqueRootHandling second?
>> 
>> If anything, I would expect the entire check of m_opaqueRoots.contains(ptr) to be only done after checking m_needsExtraOpaqueRootHandling, especially if it is UNLIKELY.
> 
> I check found first because found is already in memory.  If not found, we save on triggering a load of m_needsExtraOpaqueRootHandling.

I had missed that we eventually return found and so we need to compute it even if !m_needsExtraOpaqueRootHandling.
Comment 5 Mark Lam 2021-05-19 19:45:22 PDT 
Comment on attachment 429126 [details]
proposed patch.

Thanks for the review.  Landing now.
Comment 6 EWS 2021-05-19 20:34:36 PDT 
Committed r277773 (237934@main): <https://commits.webkit.org/237934@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 429126 [details].