[Bug tree-optimization/42720] Problematic condition simplification logic at unswitch-loops pass

[rguenth at gcc dot gnu dot org]

------- Comment #6 from rguenth at gcc dot gnu dot org  2010-01-30 10:59 -------
The interesting thing is that the empty loop is not removed by
the control-dependent DCE pass that follows the 2nd VRP.

This happens because


<bb 3>:
  if (obj_7(D) != 0B)
    goto <bb 5>;
  else
    goto <bb 4>;

<bb 4>:
  # i_36 = PHI <0(3)>

<bb 10>:
  # i_25 = PHI <i_36(4), i_35(12)>
  i_35 = i_25 + 1;
  if (i_35 != num_6(D))
    goto <bb 12>;
  else
    goto <bb 8>;

<bb 12>:
  goto <bb 10>;

...

<bb 7>:
Invalid sum of incoming frequencies 6888, should be 7735
  # s_2 = PHI <s_19(6), s_23(11)>
  i_20 = i_18 + 1;
  if (i_20 != num_6(D))
    goto <bb 13>;
  else
    goto <bb 8>;

...

<bb 8>:
  # s_3 = PHI <s_2(7), 0(10)>


the PHI node defining s_3 marks the controlling statement necessary
which in turn marks the loop necessary.  Now the question is if this
is an inherent limitation of CD-DCE or if the controlling stmt
should be that in BB3 instead (the immediate common dominator of
bb7 and bb10 is bb3).  Steven?  You implemented that CD-DCE stuff?


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org, hubicka at gcc dot gnu
                   |                            |dot org
           Keywords|                            |missed-optimization


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42720