[Webkit-unassigned] [Bug 145366] FTL is not working on Windows.

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=145366

--- Comment #6 from peavo at outlook.com ---
(In reply to comment #5)
> (In reply to comment #4)
> > (In reply to comment #3)
> > > Comment on attachment 253683 [details]
> > > Patch
> > > 
> > > View in context:
> > > https://bugs.webkit.org/attachment.cgi?id=253683&action=review
> > > 
> > > > Source/JavaScriptCore/ChangeLog:29
> > > > +        (JSC::FTL::canCompile): llvm does not support ArithPow on Windows.
> > > 
> > > This is the wrong approach. You're preventing the compilation of an entire
> > > function because of an intrinsic. Instead you should implement ArithPow on
> > > Windows using a function call to a helper.
> > 
> > Ok, thanks for reviewing :) I will look into this.
> 
> Note that you could literally call the pow() function.  It's possible that
> if you emit a call to an extern function named "pow" then LLVM might try to
> turn this into an intrinsic and then fail in the backend if the intrinsic is
> unsupported (I don't know if this is happening here, but I remember such a
> LLVM JIT bug with other math functions).  One way to avoid this bug is to
> grab a pointer to the pow() function, and then emit a call to that pointer
> constant.  We've done this a lot in other such thorny cases.

Ok, thanks for the tip :)
I also found this llvm bug: https://llvm.org/bugs/show_bug.cgi?id=8379

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20150526/aa4fba84/attachment.html>