Bug 146850 – jsc-tailcall: Implement the tail call opcodes in the DFG

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

[Basile Clement]

...

[Basile Clement]

*** Bug 147198 has been marked as a duplicate of this bug. ***

[Basile Clement]

Created attachment 257988 [details]
Patch

[Basile Clement]

Just realized this is missing callee save stuff.

[Basile Clement]

Created attachment 258116 [details]
Patch

[Michael Saboff]

Comment on attachment 258116 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=258116&action=review

r+me with a couple of comments.

> Source/JavaScriptCore/ChangeLog:13
> +         - EmulatedTailCall and EmulatedTailCallVarargs are here to perform a

What about changing Emulated* to Inlined*?

> Source/JavaScriptCore/dfg/DFGOSRExitCompilerCommon.cpp:139
> +    // FIXME: We shouldn't leave holes on the stack when performing an OSR exit

Please add a bug for the FIXME

> Source/JavaScriptCore/dfg/DFGSpeculativeJIT32_64.cpp:881
> +        m_jit.breakpoint();

This should only be for debug builds.

> Source/JavaScriptCore/dfg/DFGSpeculativeJIT64.cpp:842
> +        m_jit.breakpoint();

Ditto

[Basile Clement]

(In reply to comment #5)
> Comment on attachment 258116 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=258116&action=review
> 
> r+me with a couple of comments.
> 
> > Source/JavaScriptCore/ChangeLog:13
> > +         - EmulatedTailCall and EmulatedTailCallVarargs are here to perform a
> 
> What about changing Emulated* to Inlined*?

I find InlinedTailCall to be confusing since the tail call itself is not inlined, it is inside an inlined function. And TailCallInsideAnInlinedFunction is not great either. But I'm open to suggestions.
 
> > Source/JavaScriptCore/dfg/DFGOSRExitCompilerCommon.cpp:139
> > +    // FIXME: We shouldn't leave holes on the stack when performing an OSR exit
> 
> Please add a bug for the FIXME

Oops, thought I did that already. I'll add it.

[Michael Saboff]

(In reply to comment #6)
> (In reply to comment #5)
> > Comment on attachment 258116 [details]
> > > Source/JavaScriptCore/ChangeLog:13
> > > +         - EmulatedTailCall and EmulatedTailCallVarargs are here to perform a
> > 
> > What about changing Emulated* to Inlined*?
> 
> I find InlinedTailCall to be confusing since the tail call itself is not
> inlined, it is inside an inlined function. And
> TailCallInsideAnInlinedFunction is not great either. But I'm open to
> suggestions.

What about TailCallInlinedCaller?

[Basile Clement]

Committed r187791 <https://trac.webkit.org/changeset/187791>.