[llvm-bugs] [Bug 45470] GVN incorrectly optimizes away a non-local load

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=45470

Changpeng Fang <changpeng.fang at amd.com> changed:

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

--- Comment #3 from Changpeng Fang <changpeng.fang at amd.com> ---
Thanks for looking into this. Right the reduced test does not reproduce the
issue. So I attached the original one extract.O1.ll:

We don't expect to see "chfang(float undef)" after
opt -O1 -gvn extraxt.O1.ll -S: 


bb1355:                                           ; preds = %bb1379, %bb1346
  %tmp1356 = phi i32 [ 0, %bb1346 ], [ %tmp1383, %bb1379 ]
  %tmp1357 = shl i32 %tmp1356, 2
  %tmp1358 = getelementptr inbounds [16 x i16], [16 x i16] addrspace(5)* %tmp,
i32 0, i32 %tmp1357
  store i16 11878, i16 addrspace(5)* %tmp1358, align 2, !tbaa !28
  ...


.loopexit:                                        ; preds = %.preheader,
%.loopexit93
  %tmp1394 = bitcast [16 x i16] addrspace(5)* %tmp to half addrspace(5)*
  %tmp1395 = load half, half addrspace(5)* %tmp1394, align 2, !tbaa !18
  %tmp1396 = fpext half %tmp1395 to float
  call void @chfang(float %tmp1396)

-- 
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/20200408/8a4fb4a8/attachment.html>