[Bug rtl-optimization/19296] New: [3.3 regression] Range check on short miscompiled at -O

[Bug rtl-optimization/19296] New: [3.3 regression] Range check on short miscompiled at -O

[falk at debian dot org]

[forwarded from http://bugs.debian.org/288721]

This fails with 3.3.5 at -O and higher. Regression from 2.95. Also doesn't fail
on MIPS or Alpha. Debian 3.4.4 20041218 is OK.

void abort(void);
__attribute__((noinline)) void f(unsigned short ad) {
    if (ad >= 0x4000 && ad < 0xc000) 
        abort();
}

int main(void) {
    f(0xff00); 
    return 0;
}

-- 
           Summary: [3.3 regression] Range check on short miscompiled at -O
           Product: gcc
           Version: 3.3.5
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: critical
          Priority: P2
         Component: rtl-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: i386-linux
  GCC host triplet: i386-linux
GCC target triplet: i386-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug middle-end/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-06 19:40 -------
It also fails with last nights build of gcc of the mainline.
And 3.4.0, 3.3.3, 3.2.3 and 3.0.4.

Here is a testcase which also fails too (this is what fold does):
void abort ();
void f1(unsigned short ad)
{
        if ((short int) (ad - 16384) >= 0)
                abort ();
}
int main (void)
{
        f1(0xff00);
        return 0;
}

The RTL is wrong right away:
(insn 13 11 14 (set (reg:HI 60)
        (subreg/u:HI (reg/v:SI 58 [ ad ]) 0)) -1 (nil)
    (nil))

(insn 14 13 15 (parallel [
            (set (reg:SI 61)
                (plus:SI (subreg:SI (reg:HI 60) 0)
                    (const_int -16384 [0xffffc000])))
            (clobber (reg:CC 17 flags))
        ]) -1 (nil)
    (nil))

(insn 15 14 16 (set (reg:CCGOC 17 flags)
        (compare:CCGOC (subreg:HI (reg:SI 61) 0)
            (const_int 0 [0x0]))) -1 (nil)
    (nil))

(jump_insn 16 15 0 (set (pc)
        (if_then_else (lt (reg:CCGOC 17 flags)
                (const_int 0 [0x0]))
            (label_ref 0)
            (pc))) -1 (nil)
    (expr_list:REG_BR_PROB (const_int 9900 [0x26ac])
        (nil)))


on PPC we get:
(insn 11 9 12 (set (reg:HI 119)
        (subreg/u:HI (reg/v:SI 118 [ ad ]) 2)) -1 (nil)
    (nil))

(insn 12 11 13 (set (reg:SI 120)
        (plus:SI (subreg:SI (reg:HI 119) 0)
            (const_int -16384 [0xffffffffffffc000]))) -1 (nil)
    (nil))

(insn 13 12 14 (set (reg:SI 121)
        (sign_extend:SI (subreg:HI (reg:SI 120) 2))) -1 (nil)
    (nil))

(insn 14 13 15 (set (reg:CC 122)
        (compare:CC (reg:SI 121)
            (const_int 0 [0x0]))) -1 (nil)
    (nil))

(jump_insn 15 14 0 (set (pc)
        (if_then_else (lt (reg:CC 122)
                (const_int 0 [0x0]))
            (label_ref 0)
            (pc))) -1 (nil)
    (expr_list:REG_BR_PROB (const_int 9900 [0x26ac])
        (nil)))


Notice the sign_extend in the PPC but not in the x86 case.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
          Component|rtl-optimization            |middle-end
     Ever Confirmed|                            |1
  GCC build triplet|i386-linux                  |
   GCC host triplet|i386-linux                  |
      Known to fail|                            |3.4.0 4.0.0 3.0.4 3.2.3
                   |                            |3.3.3
      Known to work|                            |2.95.3
   Last reconfirmed|0000-00-00 00:00:00         |2005-01-06 19:40:30
               date|                            |
            Summary|[3.3 regression] Range check|[3.3/3.4/4.0 regression]
                   |on short miscompiled at -O  |Range check on short
                   |                            |miscompiled at -O
   Target Milestone|---                         |3.4.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug middle-end/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-01-06 20:37 -------
It only fails on i686, neither on i386 nor i586.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|i386-linux                  |i686-linux
      Known to fail|3.4.0 4.0.0 3.0.4 3.2.3     |3.4.0 3.4.4 4.0.0 3.0.4
                   |3.3.3                       |3.2.3 3.3.3 3.3.6
   Last reconfirmed|2005-01-06 19:40:30         |2005-01-06 20:37:10
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug middle-end/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-01-06 20:56 -------
Investigating.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |ebotcazou at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug middle-end/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-01-06 22:24 -------
 > The RTL is wrong right away:
> (insn 13 11 14 (set (reg:HI 60)
>         (subreg/u:HI (reg/v:SI 58 [ ad ]) 0)) -1 (nil)
>     (nil))
> 
> (insn 14 13 15 (parallel [
>             (set (reg:SI 61)
>                 (plus:SI (subreg:SI (reg:HI 60) 0)
>                     (const_int -16384 [0xffffc000])))
>             (clobber (reg:CC 17 flags))
>         ]) -1 (nil)
>     (nil))
> 
> (insn 15 14 16 (set (reg:CCGOC 17 flags)
>         (compare:CCGOC (subreg:HI (reg:SI 61) 0)
>             (const_int 0 [0x0]))) -1 (nil)
>     (nil))
> 
> (jump_insn 16 15 0 (set (pc)
>         (if_then_else (lt (reg:CCGOC 17 flags)
>                 (const_int 0 [0x0]))
>             (label_ref 0)
>             (pc))) -1 (nil)
>     (expr_list:REG_BR_PROB (const_int 9900 [0x26ac])
>         (nil)))

I don't think the initial RTL is wrong: note that, while we perform the addition
in SImode, we still compare in HImode.  The problem appears to come from the
combiner, which gets rid of the HImode in the comparison.

> on PPC we get:
> (insn 11 9 12 (set (reg:HI 119)
>         (subreg/u:HI (reg/v:SI 118 [ ad ]) 2)) -1 (nil)
>     (nil))
> 
> (insn 12 11 13 (set (reg:SI 120)
>         (plus:SI (subreg:SI (reg:HI 119) 0)
>             (const_int -16384 [0xffffffffffffc000]))) -1 (nil)
>     (nil))
> 
> (insn 13 12 14 (set (reg:SI 121)
>         (sign_extend:SI (subreg:HI (reg:SI 120) 2))) -1 (nil)
>     (nil))
> 
> (insn 14 13 15 (set (reg:CC 122)
>         (compare:CC (reg:SI 121)
>             (const_int 0 [0x0]))) -1 (nil)
>     (nil))
> 
> (jump_insn 15 14 0 (set (pc)
>         (if_then_else (lt (reg:CC 122)
>                 (const_int 0 [0x0]))
>             (label_ref 0)
>             (pc))) -1 (nil)
>     (expr_list:REG_BR_PROB (const_int 9900 [0x26ac])
>         (nil)))
> 
> 
> Notice the sign_extend in the PPC but not in the x86 case.

This RTL is indeed more elegant and probably less error-prone.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug rtl-optimization/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-01-07 00:46 -------
Confirmed, it's the combiner optimistically getting rid of the HImode subreg.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |rtl-optimization


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug rtl-optimization/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[debian-gcc at lists dot debian dot org]

------- Additional Comments From debian-gcc at lists dot debian dot org  2005-01-08 16:07 -------
[add reference to http://bugs.debian.org/288721]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |debian-gcc at lists dot
                   |                            |debian dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug rtl-optimization/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-09 02:12 -------
Patch here: <http://gcc.gnu.org/ml/gcc-patches/2005-01/msg00464.html>.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug rtl-optimization/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-01-18 08:26 -------
Subject: Bug 19296

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	ebotcazou@gcc.gnu.org	2005-01-18 08:26:21

Modified files:
	gcc            : ChangeLog combine.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: short-compare-1.c short-compare-2.c 

Log message:
	PR rtl-optimization/19296
	* combine.c (simplify_comparison): Rewrite the condition under
	which a non-paradoxical SUBREG of a PLUS can be lifted when
	compared against a constant.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.7168&r2=2.7169
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/combine.c.diff?cvsroot=gcc&r1=1.466&r2=1.467
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.4901&r2=1.4902
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/short-compare-1.c.diff?cvsroot=gcc&r1=NONE&r2=1.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/short-compare-2.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug rtl-optimization/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-01-18 08:33 -------
Subject: Bug 19296

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_4-branch
Changes by:	ebotcazou@gcc.gnu.org	2005-01-18 08:33:28

Modified files:
	gcc            : ChangeLog combine.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: short-compare-1.c short-compare-2.c 

Log message:
	PR rtl-optimization/19296
	* combine.c (simplify_comparison): Rewrite the condition under
	which a non-paradoxical SUBREG of a PLUS can be lifted when
	compared against a constant.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=2.2326.2.775&r2=2.2326.2.776
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/combine.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.400.4.11&r2=1.400.4.12
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.3389.2.344&r2=1.3389.2.345
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/short-compare-1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=NONE&r2=1.1.2.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/short-compare-2.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=NONE&r2=1.1.2.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug rtl-optimization/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-01-18 08:39 -------
Subject: Bug 19296

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_3-branch
Changes by:	ebotcazou@gcc.gnu.org	2005-01-18 08:39:30

Modified files:
	gcc            : ChangeLog combine.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: short-compare-1.c short-compare-2.c 

Log message:
	PR rtl-optimization/19296
	* combine.c (simplify_comparison): Rewrite the condition under
	which a non-paradoxical SUBREG of a PLUS can be lifted when
	compared against a constant.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.16114.2.1052&r2=1.16114.2.1053
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/combine.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.325.2.18&r2=1.325.2.19
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.2261.2.395&r2=1.2261.2.396
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/short-compare-1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=NONE&r2=1.1.4.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/short-compare-2.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=NONE&r2=1.1.4.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296

[Bug rtl-optimization/19296] [3.3/3.4/4.0 regression] Range check on short miscompiled at -O

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2005-01-18 08:41 -------
Patch applied.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|3.4.4                       |3.3.6


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19296