[llvm-bugs] [Bug 25808] [InstCombine] canonicalize extractelement reductions with shuffles

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=25808

Sanjay Patel <spatel+llvm at rotateright.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|CONFIRMED                   |RESOLVED
         Resolution|---                         |FIXED

--- Comment #9 from Sanjay Patel <spatel+llvm at rotateright.com> ---
(In reply to Simon Pilgrim from comment #8)
> Resolve this? Reduction costs have been tidied up a lot in the past year,
> allowing the SLP to vectorize this to a reduction pattern, even on sse2
> targets.

Yes, that's probably better than trying to do this in instcombine. We're still
unable to create arbitrary shuffles in instcombine because we don't know what
kind of target support there exists to back that up.

For x86, I think we get the expected codegen for all targets, including a
horizontal math op if that's desirable.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20190110/a86bbd3f/attachment-0001.html>