This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug c/16632] New: A couple fold-const.c optimizations are non-functional


In fold-const.c there are a couple of bitmask/comparision optimizations:
      /* If we have (A & C) == D where D & ~C != 0, convert this into 0.
and
      /* If we have (A | C) == D where C & ~D != 0, convert this into 0.

Sample C code:
  void subr (int i)
  {
    if ((i|3)  == 1) return; /* never taken */
    if ((i&4)  == 2) return; /* never taken */
  }


Unfortunately fold-const.c neglects to fold ~C (~D) and so
the detection of these cases is non-functional.
I'm not sure how to demonstrate this outside the debugger,
since later pieces of gcc seem to clean this up.

For example ~D is computed on approx line 8391 with

                            build1 (BIT_NOT_EXPR, TREE_TYPE (arg1), arg1)));

This should be fold(build1 ...), or fold_not_const(arg1, TREE_TYPE(arg1)).
Similarly for ~C on approx line 8373.

-- 
           Summary: A couple fold-const.c optimizations are non-functional
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Severity: minor
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: trt at acm dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16632


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]