[llvm-bugs] [Bug 26943] New: InstCombine: frem X, (C ? 0 : Y) -> frem X, Y changes result

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=26943

            Bug ID: 26943
           Summary: InstCombine: frem X, (C ? 0 : Y) -> frem X, Y changes
                    result
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: MacOS X
            Status: NEW
          Severity: normal
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: andres.noetzli at gmail.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

The following program:

$ cat ../example.ll
define i1 @main() {
  %w = call float @foo(i1 1, float 17.0)
  %w2 = frem float 17.0, 42.0
  %rr = fcmp oeq float %w, %w2
  ret i1 %rr
}

define float @foo(i1 %c, float %x) {
  %r = select i1 %c, float 0.0, float 42.0
  %rr = frem float %x, %r
  ret float %rr
}

gets optimized to:

$ bin/opt < ../example.ll -instcombine -S > ../example2.ll && cat
../example2.ll
; ModuleID = '<stdin>'

define i1 @main() {
  %w = call float @foo(i1 true, float 1.700000e+01)
  %rr = fcmp oeq float %w, 1.700000e+01
  ret i1 %rr
}

define float @foo(i1 %c, float %x) {
  %rr = frem float %x, 4.200000e+01
  ret float %rr
}

The output of the original program and the optimized program is not the same:

$ bin/lli ../example.ll; echo $?
0
$ bin/lli ../example2.ll; echo $?
1

It seems like the optimization assumes that a frem %x, 0.0 results in undefined
behavior but according to [0], the result should just be NaN and an error flag
should be set. My understanding is that frem corresponds to C's fmod (see e.g.
[1]), so it seems like the optimization is invalid.

[0] http://pubs.opengroup.org/onlinepubs/9699919799/functions/fmod.html
[1]
http://llvm.org/docs/doxygen/html/classllvm_1_1APFloat.html#ab715d4ba706dd8dd1c7d51e43d067d4c

-- 
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/20160314/ef5b8faa/attachment.html>