Bug 154897 – [[SetPrototypeOf]] should be a fully virtual method in ClassInfo::methodTable

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Saam Barati]

We can probably make this fast just by using the setPrototypeOfInline paradigm
where we will often be able to inline this call as JSObject::setPrototypeWithCycleCheck.

This is needed for Proxy

[Saam Barati]

Created attachment 272677 [details]
perf results

looks neutral

[Saam Barati]

Created attachment 272678 [details]
patch

[Saam Barati]

landed in:
http://trac.webkit.org/changeset/197467

[Geoffrey Garen]

Comment on attachment 272678 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=272678&action=review

> Source/JavaScriptCore/runtime/ClassInfo.h:113
> +    typedef bool (*SetPrototypeOfFunctionPtr)(JSObject*, ExecState*, JSValue);
> +    SetPrototypeOfFunctionPtr setPrototypeOf;

When referring to the object constructor, the spec says "setPrototypeOf" because the object constructor sets the prototype *of* another object.

But when we refer to an object setting its own prototype, that should be called "setPrototype". This applies to lots of places in this patch.

> Source/JavaScriptCore/runtime/JSObject.h:122
> +    ALWAYS_INLINE bool setPrototypeOfInline(VM& vm, ExecState* exec, JSValue prototype)

Why do we have this inline variant? I think you should remove it.

[Saam Barati]

(In reply to comment #4)
> Comment on attachment 272678 [details]
> patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=272678&action=review
> 
> > Source/JavaScriptCore/runtime/ClassInfo.h:113
> > +    typedef bool (*SetPrototypeOfFunctionPtr)(JSObject*, ExecState*, JSValue);
> > +    SetPrototypeOfFunctionPtr setPrototypeOf;
> 
> When referring to the object constructor, the spec says "setPrototypeOf"
> because the object constructor sets the prototype *of* another object.
> 
> But when we refer to an object setting its own prototype, that should be
> called "setPrototype". This applies to lots of places in this patch.
> 
> > Source/JavaScriptCore/runtime/JSObject.h:122
> > +    ALWAYS_INLINE bool setPrototypeOfInline(VM& vm, ExecState* exec, JSValue prototype)
> 
> Why do we have this inline variant? I think you should remove it.

Follow up to your comment in:
http://trac.webkit.org/changeset/197484