Bug 164600 – Graph::methodOfGettingAValueProfileFor() should be returning the profile for the operand node.

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

[Mark Lam]

Currently, Graph::methodOfGettingAValueProfileFor() assumes that the operand DFG node that it is provided with always has a different origin than the node that is using that operand.  For example, in a DFG graph that looks like this:
    a: ...
    b: ArithAdd(@a, ...)

... when emitting speculation checks on @a for the ArithAdd node at @b, Graph::methodOfGettingAValueProfileFor() is passed @a, and expects @a's to originate from a different bytecode than @b.

However, op_negate can be compiled into the following series of nodes:

    a: BooleanToNumber(...)
    b: DoubleRep(@a)
    c: ArithNegate(@b)

All 3 nodes maps to the same op_negate bytecode i.e. they have the same origin.  When the speculativeJIT does a speculationCheck in DoubleRep, it calls Graph::methodOfGettingAValueProfileFor() to get the ArithProfile for the BooleanToNumber node.  But because all 3 nodes have the same origin, Graph::methodOfGettingAValueProfileFor() erroneously returns the ArithProfile for the op_negate.  Subsequently, the OSR exit ramp will modify the ArithProfile of the op_negate and corrupt its profile.  Instead, what the OSR exit ramp should be doing is update the ArithProfile of op_negate's operand i.e. BooleanToNumber's operand in this case.

The fix is to always pass the current node we're generating code for (in addition to the operand node) to Graph::methodOfGettingAValueProfileFor().  This way, we can determine if the profile is valid if and only if the current node and its operand node does not have the same origin.

[Mark Lam]

<rdar://problem/28828676>

[Mark Lam]

Created attachment 294385 [details]
proposed patch.

[Filip Pizlo]

Comment on attachment 294385 [details]
proposed patch.

Thank you for cleaning this up!

[Geoffrey Garen]

Seems correct to me. We probably need Phil to review this.

> Source/JavaScriptCore/dfg/DFGGraph.cpp:1619
> +    Node* node = operandNode;
>      while (node) {

I think this would be clearer as: for (Node* node = operandNode; node; )

[Mark Lam]

(In reply to comment #4)
> Seems correct to me. We probably need Phil to review this.
> 
> > Source/JavaScriptCore/dfg/DFGGraph.cpp:1619
> > +    Node* node = operandNode;
> >      while (node) {
> 
> I think this would be clearer as: for (Node* node = operandNode; node; )

Thanks.  I'll apply this before landing.

[Mark Lam]

Thanks for the reviews.  Landed in r208560: <http://trac.webkit.org/r208560>.