[llvm-bugs] [Bug 38150] New: X86InstrInfo::analyzeBranch doesn't preserve undef/kill flags on $eflags

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=38150

            Bug ID: 38150
           Summary: X86InstrInfo::analyzeBranch doesn't preserve
                    undef/kill flags on $eflags
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: francisvm at yahoo.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 20551
  --> https://bugs.llvm.org/attachment.cgi?id=20551&action=edit
branchfolding-operands.mir

llc -run-pass branch-folder will transform this:

> bb.0:
>    JE_1 %bb.1, implicit undef $eflags
>    JMP_1 %bb.2
>  bb.1:
>    RET 2, undef $eax
>  bb.2:
>    RET 0, undef $eax

into

> bb.0:
>   JNE_1 %bb.2, implicit $eflags
> bb.1:
>   RET 2, undef $eax
> bb.2:
>   RET 0, undef $eax

It drops the undef from the $eflags which causes the machine verifier to
complain.

I tracked this down to:

* when X86InstrInfo::analyzeBranchImpl changes the branch to its opposite
because it only copies the MBB operand
* when BranchFolding::TailMergeBlocks calls removeBranch then insertBranch.
Analyze branch doesn’t keep the eflags operand in the Cond vector returned by
analyze branch, so insertBranch won’t be able to re-insert it with the correct
operands.

Attached MIR test.

-- 
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/20180712/86efd5e2/attachment.html>