[llvm-bugs] [Bug 48832] New: Miscompile with -opt -loop-vectorize after 8871a4b4cab8

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=48832

            Bug ID: 48832
           Summary: Miscompile with -opt -loop-vectorize after
                    8871a4b4cab8
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Loop Optimizer
          Assignee: unassignedbugs at nondot.org
          Reporter: mikael.holmen at ericsson.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 24405
  --> https://bugs.llvm.org/attachment.cgi?id=24405&action=edit
bbi-51947.ll reproducer

With
 opt -mtriple=s390x-unknown-linux -mcpu=z13 -loop-vectorize -S -o -
bbi-51947.ll

I get a
 store <4 x i32> poison, <4 x i32>* %6, align 1

which I htink is incorrect.

This starts happening with 8871a4b4cab8:


    [Constant] Update ConstantVector::get to return poison if all input elems
are poison

    The diff was reviewed at D93994

Before the patch I got

store <4 x i32> zeroinitializer, <4 x i32>* %6, align 1

The input is

%arrayt = type [64 x i32]

@v_146 = external global %arrayt, align 1

define void @foo() {
entry:
  br label %for.cond

for.cond:                                         ; preds = %cond.end, %entry
  %storemerge = phi i16 [ 0, %entry ], [ %inc, %cond.end ]
  %cmp = icmp slt i16 %storemerge, 15
  br i1 %cmp, label %for.body, label %for.end

for.body:                                         ; preds = %for.cond
  br i1 true, label %cond.false, label %land.rhs

land.rhs:                                         ; preds = %for.body
  br i1 poison, label %cond.end, label %cond.false

cond.false:                                       ; preds = %for.body,
%land.rhs
  br label %cond.end

cond.end:                                         ; preds = %land.rhs,
%cond.false
  %cond = phi i32 [ 0, %cond.false ], [ 1, %land.rhs ]
  %arrayidx = getelementptr inbounds %arrayt, %arrayt* @v_146, i16 0, i16
%storemerge
  store i32 %cond, i32* %arrayidx, align 1
  %inc = add nsw i16 %storemerge, 1
  br label %for.cond

for.end:                                          ; preds = %for.cond
  ret void
}

Note the

for.body:                                         ; preds = %for.cond
  br i1 true, label %cond.false, label %land.rhs

land.rhs:                                         ; preds = %for.body
  br i1 poison, label %cond.end, label %cond.false

%land.rhs is dead due to the "br i1 true" branch, but I think it still is
involved somehow in making the stored value poison with this patch.

-- 
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/20210121/6fa08b8e/attachment.html>