[llvm-bugs] [Bug 41918] New: incorrect code generated in llvm Early CSE w/ MemorySSA phase

via llvm-bugs llvm-bugs at lists.llvm.org
Thu May 16 22:15:57 PDT 2019


https://bugs.llvm.org/show_bug.cgi?id=41918

            Bug ID: 41918
           Summary: incorrect code generated in llvm Early CSE w/
                    MemorySSA phase
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: ys114321 at gmail.com
                CC: htmldeveloper at gmail.com, llvm-bugs at lists.llvm.org

Created attachment 21959
  --> https://bugs.llvm.org/attachment.cgi?id=21959&action=edit
test.sh

The original issue is reported at bcc mailing list.
  https://github.com/iovisor/bcc/issues/2352

Incorrect code is generated for a bpf C program.

After some analysis, it looks the error in Early CSE w/ MemorySSA pass.
Before:

  %27 = call i32 inttoptr (i64 4 to i32 (i8*, i64, i8*)*)(i8* nonnull %13, i64
65, i8* nonnull %26) #6, !dbg !1627 
  %28 = getelementptr inbounds [65 x i8], [65 x i8]* %3, i64 0, i64 0, !dbg
!1628 
  call void @llvm.lifetime.end.p0i8(i64 65, i8* nonnull %13) #6, !dbg !1629 
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* nonnull align 16 %10, i8* nonnull
align 16 %28, i64 64, i1 false), !dbg !1630

After

  %25 = call i32 inttoptr (i64 4 to i32 (i8*, i64, i8*)*)(i8* nonnull %11, i64
65, i8* nonnull %24) #6, !dbg !1627
  call void @llvm.lifetime.end.p0i8(i64 65, i8* nonnull %11) #6, !dbg !1628
  call void @llvm.memcpy.p0i8.p0i8.i64(i8* nonnull align 16 %10, i8* nonnull
align 16 %11, i64 64, i1 false), !dbg !1629

After Early CSE pass, the variable %11 is marked as lifetime.end, but it still
used afterwards. Eventually wrong code is generated.

Based on the bcc issue, the bug exists in earlier llvm versions as well.

I have created a stand alone test case, test.sh and bug.i. Run through the code
and check the IR, you will see the incorrect IR generated.

-- 
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/20190517/33c3ce03/attachment.html>


More information about the llvm-bugs mailing list