[Bug tree-optimization/25680] Store CCP does not understand REALPART_EXPR < COMPLEX_CST >

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2006-01-05 15:25 -------
The testcase is comment #0.

In .final_cleanup:
  t = __complex__ (0, 0);
  REALPART_EXPR <t> = REALPART_EXPR <t> + 2;
  IMAGPART_EXPR <t> = IMAGPART_EXPR <t> + 2;
  D.1520 = g (&t);
  return D.1520;


------
in .store_ccp:
Visiting statement:
#   SFT.0_3 = V_MUST_DEF <SFT.0_1>;
#   SFT.1_4 = V_MUST_DEF <SFT.1_2>;
t = __complex__ (0, 0);

Lattice value changed to CONSTANT __complex__ (0, 0).  Adding SSA edges to
worklist.
Lattice value changed to CONSTANT __complex__ (0, 0).  Adding SSA edges to
worklist.

Visiting statement:
#   VUSE <SFT.1_4>;
D.1516_5 = REALPART_EXPR <t>;

Lattice value changed to VARYING.  Adding SSA edges to worklist.


I should note that on powerpc-darwin we don't even get the constant prop at the
RTL level either.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25680