[Bug tree-optimization/16220] New: Jump pessimization

[Bug tree-optimization/16220] New: Jump pessimization

[falk at debian dot org]

long ticks(long val, long high, long prev) {
    return high + ((long) (val < prev) << 32);
}

gcc 3.3 produces this nice branch-free code:

cmplt   a0,a2,v0
sll     v0,0x20,t1
addq    a1,t1,v0

whereas mainline introduces a branch and pointless constants:

	cmplt   a0,a2,a2
	clr     v0
	beq     a2,L14
	lda     t0,1
	sll     t0,0x20,v0
L14:	addq    v0,a1,v0
	ret

-- 
           Summary: Jump pessimization
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: falk at debian dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: alphaev68-unknown-linux-gnu
  GCC host triplet: alphaev68-unknown-linux-gnu
GCC target triplet: alphaev68-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220

[Bug tree-optimization/16220] [3.5 Regression] Jump pessimization

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |3.5.0
      Known to work|                            |3.3
            Summary|Jump pessimization          |[3.5 Regression] Jump
                   |                            |pessimization
   Target Milestone|---                         |3.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220

[Bug tree-optimization/16220] [3.5 Regression] Jump pessimization

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-06-26 19:46 -------
Confirmed, the problem comes the fact PHI-OPT not being able to optimize this and something 
converting (a<b) to (a<b)?0:1 where the 1 is overflowed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-06-26 19:46:23
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220

[Bug tree-optimization/16220] [3.5 Regression] Jump pessimization

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-07-17 22:52 -------
Actually what has happened now is that (long) (val < prev) << 32 is changed into a < b ? 1 <<32 : 0. :(
So what needs to happen is PHI-OPT needs to detect this situational.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2004-06-26 19:46:23         |2004-07-17 22:52:49
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220

[Bug tree-optimization/16220] [4.0 Regression] Jump pessimization

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-09-30 18:06 -------
Roger I almost think this was caused by:
2004-05-20  Roger Sayle  <roger@eyesopen.com>

        PR middle-end/3074
        * fold-const.c (strip_compound_expr): Delete function.
        (count_cond): Delete function.
        (fold_binary_op_with_conditional_arg): Only perform transformations
        "a + (b?c:d) -> b ? a+c : a+d" and "(b?c:d) + a -> b ? c+a : d+a"
        when a is constant.  This greatly simplifies this routine.
but I could be wrong.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |roger at eyesopen dot com
      Known to fail|4.0                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220

[Bug tree-optimization/16220] [4.0 Regression] Jump pessimization

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |minor
  GCC build triplet|alphaev68-unknown-linux-gnu |
   GCC host triplet|alphaev68-unknown-linux-gnu |
 GCC target triplet|alphaev68-unknown-linux-gnu |alpha*-*-*


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220

[Bug tree-optimization/16220] [4.0 Regression] Jump pessimization

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-04 19:25 -------
Patch here: <http://gcc.gnu.org/ml/gcc-patches/2005-01/msg00210.html>.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220

[Bug tree-optimization/16220] [4.0 Regression] Jump pessimization

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-10 01:05 -------
Fixed by:
2005-01-08  Roger Sayle  <roger@eyesopen.com>

        * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
        the latent bug in rtl_delete_block has been resolved.
aka
2005-01-04  Roger Sayle  <roger@eyesopen.com>

        * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
        if deleting the then-block allows the test-block to fallthru to the
        else-block.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16220