[llvm-bugs] [Bug 45778] New: InstSimplify: fadd (nsz op), +0 incorrectly removed

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=45778

            Bug ID: 45778
           Summary: InstSimplify: fadd (nsz op), +0 incorrectly removed
           Product: libraries
           Version: trunk
          Hardware: All
                OS: All
            Status: NEW
          Keywords: miscompilation
          Severity: normal
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: nunoplopes at sapo.pt
                CC: llvm-bugs at lists.llvm.org, regehr at cs.utah.edu,
                    spatel+llvm at rotateright.com

Test: Transforms/InstSimplify/fast-math.ll
Summary: +0 + +0 == +0 and -0 + +0 == +0 in default rounding mode. Hence below
when %nsz is +/-0 the function returns +0 only, while the optimized function
returns +/-0.

define float @fold_fadd_cannot_be_neg0_nsz_src_x_0(float %a, float %b) {
  %nsz = fmul nsz float %a, %b
  %add = fadd float %nsz, 0.000000
  ret float %add
}
=>
define float @fold_fadd_cannot_be_neg0_nsz_src_x_0(float %a, float %b) {
  %nsz = fmul nsz float %a, %b
  ret float %nsz
}
Transformation doesn't verify!
ERROR: Value mismatch

Example:
float %a = #x01e21080 (0.000000000000?)
float %b = #x00225d01 (0.000000000000?)

Source:
float %nsz = #x80000000 (-0.0)
float %add = #x00000000 (+0.0)

Target:
float %nsz = #x80000000 (-0.0)
Source value: #x00000000 (+0.0)
Target value: #x80000000 (-0.0)

https://web.ist.utl.pt/nuno.lopes/alive2/index.php?hash=1ede71ade7988ad1&test=Transforms%2FInstSimplify%2Ffast-math.ll

-- 
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/20200502/74157161/attachment.html>