[llvm-bugs] [Bug 24288] Regression due to GMR changes in r242512

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=24288

Michael Kuperstein <michael.m.kuperstein at intel.com> changed:

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

--- Comment #8 from Michael Kuperstein <michael.m.kuperstein at intel.com> ---
Some of the regressions in eembc were fixed, but others still remain.

There are two remaining issues I can see:

1) Easy: isNonEscapingGlobalNoAlias(), for the "load from global variable"
case, does not call GetUnderlyingObject() on the pointer operand. This means it
misses cases where the load is from a GEP/BitCast ConstantExpr based on a GV.

2) Harder: this is tangentially related to PR22475. The canonicalization from
r226781 produces inttoptr casts that isNonEscapingGlobalNoAlias() can't (and
probably shouldn't try to) look through. I think the fix for this should be on
the InstCombine end, not GMR.

-- 
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/20150816/902d9b27/attachment-0001.html>