[llvm-bugs] [Bug 43711] [AMDGPU] fneg optimized to xor with sign bit removed

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=43711

Matt Arsenault <Matthew.Arsenault at amd.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |INVALID

--- Comment #2 from Matt Arsenault <Matthew.Arsenault at amd.com> ---
IEEE-754 2008 5.5.1 "Sign bit operations" states "they only affect the sign
bit" and
 "These operations may propagate non-canonical encodings."

I would not expect nir_op_fneg to behave any differently, but if you for some
reason defined it that way, emitting the separate canonicalize would be the
correct way to map it to llvm semantics. I suspect you're really missing a
canonicalize somewhere else though.

-- 
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/20191018/eb7b21f4/attachment.html>