[llvm-bugs] [Bug 47322] New: Severe optimizer regression with LLVM 11.0.0-rc2

via llvm-bugs llvm-bugs at lists.llvm.org
Wed Aug 26 15:25:36 PDT 2020


https://bugs.llvm.org/show_bug.cgi?id=47322

            Bug ID: 47322
           Summary: Severe optimizer regression with LLVM 11.0.0-rc2
           Product: libraries
           Version: 11.0
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: release blocker
          Priority: P
         Component: Interprocedural Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: kinke at gmx.net
                CC: llvm-bugs at lists.llvm.org

Created attachment 23896
  --> https://bugs.llvm.org/attachment.cgi?id=23896&action=edit
unoptimized input IR

Hi, not sure whether I've chosen the right bug component, but anyway: I've hit
an apparent severe regression while testing LLVM 11-rc2 with LDC, on both Linux
and macOS x86_64. I've attached the unoptimized .ll produced by LDC.

With LLVM 10, `opt -O3 -mtriple=x86_64-linux-gnu` yields the expected:
---
define i32 @main() local_unnamed_addr #0 {
forbody.preheader.i:
  ret i32 0
}
---

while LLVM 11-rc2 comes up with an *undefined* branch:
---
define i32 @main() local_unnamed_addr #0 {
forbody.i:
  br i1 undef, label %assertPassed, label %forbody.i.2

assertPassed:                                     ; preds = %forbody.i
  ret i32 0

forbody.i.2:                                      ; preds = %forbody.i
  tail call void @__assert(i8* getelementptr inbounds ([55 x i8], [55 x i8]*
@.str.4, i64 0, i64 0), i8* getelementptr inbounds ([55 x i8], [55 x i8]*
@.str.5, i64 0, i64 0), i32 2432) #2
  unreachable
}
---

-- 
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/20200826/abe9a50a/attachment-0001.html>


More information about the llvm-bugs mailing list