[Bug rtl-optimization/24418] New: Redudant compare to zero not eliminated

[Bug rtl-optimization/24418] New: Redudant compare to zero not eliminated

[evancheng at gcc dot gnu dot org]

Compile the attached code with -O3. It produces an redundent cmpl following a
subl:

        subl    28(%ebp), %eax
        cmpl    $0, %eax 
        je      L12
        jg      L10

Looks like it's missing a insn pattern so  combine would combine subl + cmpl.

See http://gcc.gnu.org/ml/gcc/2005-10/msg00347.html


-- 
           Summary: Redudant compare to zero not eliminated
           Product: gcc
           Version: 4.0.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: evancheng at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24418

[Bug rtl-optimization/24418] Redudant compare to zero not eliminated

[evancheng at gcc dot gnu dot org]

------- Comment #1 from evancheng at gcc dot gnu dot org  2005-10-18 00:02 -------
Created an attachment (id=10014)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=10014&action=view)
C test case


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24418

[Bug rtl-optimization/24418] Redudant compare to zero not eliminated

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2005-10-18 00:07 -------
This is a partial regression from 2.95.3 and 3.3.3, 3.2.3 and 3.0.4:
where we had:
        subl 28(%ebp),%eax
        jne .L7
...
.L7:
        testl %eax,%eax
        jle .L8


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
  GCC build triplet|i686-pc-linux-gnu           |
   GCC host triplet|i686-pc-linux-gnu           |
           Keywords|                            |missed-optimization
   Last reconfirmed|0000-00-00 00:00:00         |2005-10-18 00:07:31
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24418

[Bug rtl-optimization/24418] Redudant compare to zero not eliminated

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2005-10-18 00:09 -------
and yes eax/reg 68 (or n) is dead after the compare in combine:
(insn 37 36 39 3 (parallel [
            (set (reg/v:SI 68 [ n ])
                (minus:SI (reg:SI 71)
                    (reg/v:SI 66 [ med ])))
            (clobber (reg:CC 17 flags))
        ]) 172 {*subsi_1} (insn_list:REG_DEP_TRUE 36 (nil))
    (expr_list:REG_DEAD (reg:SI 71)
        (expr_list:REG_UNUSED (reg:CC 17 flags)
            (nil))))

(insn 39 37 40 3 (set (reg:CC 17 flags)
        (compare:CC (reg/v:SI 68 [ n ])
            (const_int 0 [0x0]))) 2 {*cmpsi_1_insn} (insn_list:REG_DEP_TRUE 37
(nil))
    (expr_list:REG_DEAD (reg/v:SI 68 [ n ])
        (nil)))


Note the REG_DEAD, so what RTH said is wrong.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24418

[Bug target/24418] Redudant compare to zero not eliminated

[evancheng at gcc dot gnu dot org]

------- Comment #4 from evancheng at gcc dot gnu dot org  2005-10-25 22:55 -------
Closing this. It's not a bug. It's not safe to eliminate the cmpl. Its result
are read by je and jg:

je reads ZF, jg reads ZF and OF.  Now consider

subl 0x80000000 0x1   (OF <- 1)
cmpl 0x7fffffff               (OF <- 0)


-- 

evancheng at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24418