[Bug middle-end/39757] inconsistency in array bounds checking with -O3

["pinskia at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39757

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-02-02 20:41:15 UTC ---
We have:
  short last[3] = { 16,16,16 }, mul[3], buf[3][3][386];
  for (i=0; i < sizeof(buf)/sizeof(short); i++)
    buf[0][0][i] = 2048;

Obviously this where one of the warning happens.

Likewise for:
  float cfilt=0, ddft[3][3][2], ppm[3][3][3];


>For one, the compiler considers it an issues that array cam_xyz[4][3] is
accessed as cam_xyz[0][j], 0 <= j < 12. This issue is found with -O3 but not
with -O2.

That warning is correct.  Now for the other part, this warning really depends
on optimization and is not always 100% of producing all the warnings it could.