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 tree-optimization/21485] [4.2/4.3/4.4 Regression] missed load PRE



------- Comment #30 from rguenth at gcc dot gnu dot org  2008-10-03 23:54 -------
Comment #6 still applies.  On the trunk we do not fully exploit the partial
redundant load of array[k] in

      if (k < j) 
        if (array[k] < array[k + 1L]) 
          ++k; 
      if (array[i] < array[k]) 

but we transform the above to

      if (k < j)
        {
          tmp1 = array[k];
          tmp2 = array[k+1];
          if (tmp1 < tmp2)
            ++k;
          else
            tmp1 = array[k+1];
        }
      if (array[i] < tmp1)

missing the full redundancy of array[k+1].

Reduced testcase:

long
NumSift (long *array, unsigned long k)
{
  if (array[k] < array[k + 1L])
    ++k;
  return array[k];
}

with integer k it gets somewhat more complicated even.

The question is whether this explains the slowdown compared to GCC 3.4.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2006-07-05 09:23:46         |2008-10-03 23:54:40
               date|                            |
            Summary|[4.2/4.3/4.4 Regression]    |[4.2/4.3/4.4 Regression]
                   |codegen regression due to   |missed load PRE
                   |PRE increasing register     |
                   |pressure (missing load PRE  |
                   |really)                     |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21485


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