Bug 221166 – Make check for full HashTables opt-in

• Status: RESOLVED
• Resolution: FIXED
• Component: Web Template Framework

[Tadeu Zagallo]

It's too expensive to always perform the check

[Tadeu Zagallo]

<rdar://problem/70902458>

[Tadeu Zagallo]

Created attachment 418786 [details]
Patch

[Saam Barati]

Comment on attachment 418786 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=418786&action=review

> Source/WTF/ChangeLog:9
> +        Having the check always on was a regression on Speedometer2.

you should say what you did, e.g, opt in metaallocator

[Yusuke Suzuki]

Comment on attachment 418786 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=418786&action=review

r=me too with one comment.

> Source/WTF/wtf/MetaAllocator.h:197
> +    HashMap<FreeSpacePtr, FreeSpaceNode*, DefaultHash<FreeSpacePtr>, HashTraits<FreeSpacePtr>, HashTraits<FreeSpaceNode*>, true> m_freeSpaceStartAddressMap;
> +    HashMap<FreeSpacePtr, FreeSpaceNode*, DefaultHash<FreeSpacePtr>, HashTraits<FreeSpacePtr>, HashTraits<FreeSpaceNode*>, true> m_freeSpaceEndAddressMap;

Some enum would be better for readability, like,

enum class HashTableHardeningMode : uint8_t {
    None,
    AssertNotFull,
};

And HashMap<FreeSpacePtr, FreeSpaceNode*, DefaultHash<FreeSpacePtr>, HashTraits<FreeSpacePtr>, HashTraits<FreeSpaceNode*>, HashTableHardeningMode:: AssertNotFull>

[Tadeu Zagallo]

Created attachment 418793 [details]
Patch for landing

[EWS]

Committed r272095: <https://trac.webkit.org/changeset/272095>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 418793 [details].