[Bug tree-optimization/100503] New: A possible divide by zero problem in function do_rpo_vn

[Bug tree-optimization/100503] New: A possible divide by zero problem in function do_rpo_vn

[yguoaz at cse dot ust.hk]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100503

            Bug ID: 100503
           Summary: A possible divide by zero problem in function
                    do_rpo_vn
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: yguoaz at cse dot ust.hk
  Target Milestone: ---

In gcc/tree-ssa-sccvn.c, the function do_rpo_vn has the following code (link to
the code location:
https://github.com/gcc-mirror/gcc/blob/releases/gcc-11/gcc/tree-ssa-sccvn.c#L7815-#L7842)

static unsigned
do_rpo_vn (function *fn, edge entry, bitmap exit_bbs,
           bool iterate, bool eliminate) {
    ...
    int nex = 0;
    ...
    for (int i = 0; i < n; ++i) {
        basic_block bb = BASIC_BLOCK_FOR_FN (fn, rpo[i]);
        if (bb->flags & BB_EXECUTABLE)
            nex++;
    }
    ...
    statistics_histogram_event (cfun, "RPO iterations", 10*nblk / nex);
}

In the loop, the code counts the number of basic blocks with BB_EXECUTABLE flag
in variable nex and use it as divisor after exiting the loop. If no basic block
has such flag, then we will have a divide by zero problem. Is this possible ?
Or we must have at least one bb with the flag BB_EXECUTABLE ?

[Bug tree-optimization/100503] A possible divide by zero problem in function do_rpo_vn

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100503

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
                 CC|                            |rguenth at gcc dot gnu.org
   Last reconfirmed|                            |2021-05-10

[Bug tree-optimization/100503] A possible divide by zero problem in function do_rpo_vn

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100503

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
             Status|NEW                         |RESOLVED

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
There will always be at least one BB which has its BB_EXECUTABLE set.
As shown by (on line 7610):
  entry->dest->flags |= BB_EXECUTABLE;

It is hard to do this kind of static anlysis without understanding the code
fully.