[Bug rtl-optimization/23098] New: [4.1 Regression] store of 0.0 to float

[Bug rtl-optimization/23098] New: [4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

Take the following program:
int g (float* g1)
{
  *g1 = 0;
}

Running it on ppc-darwin on the mainline (with -static), we get:
LC0:
        .long   0
        .text
        .align 2
        .globl _g
_g:
        lis r2,ha16(LC0)
        lfs f0,lo16(LC0)(r2)
        stfs f0,0(r3)
        blr

On 4.0.0, we get:
_g:
        li r0,0
        stw r0,0(r3)
        blr

x86 have the same problem but only with -fPIC but that is not a regression:
.LC0:
        .long   0
        .text
        .p2align 4,,15
.globl g
        .type   g, @function
g:
        call    __i686.get_pc_thunk.cx
        addl    $_GLOBAL_OFFSET_TABLE_, %ecx
        movl    4(%esp), %eax
        movl    .LC0@GOTOFF(%ecx), %edx
        movl    %edx, (%eax)
        ret

-- 
           Summary: [4.1 Regression] store of 0.0 to float
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: minor
          Priority: P2
         Component: rtl-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pinskia at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
GCC target triplet: powerpc-darwin


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-27 18:49 -------
Actually I take back x86 is not a regression.  It is a regression from 3.2.3

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |3.3.3 3.4.0 4.0.0 4.1.0
      Known to work|                            |2.95.3 3.2.3
            Summary|[4.1 Regression] store of   |[3.4/4.0/4.1 Regression]
                   |0.0 to float                |store of 0.0 to float


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|powerpc-darwin              |powerpc-darwin, i?86-*-*
   Target Milestone|---                         |4.0.2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[dje at gcc dot gnu dot org]

------- Additional Comments From dje at gcc dot gnu dot org  2005-07-27 19:11 -------
On PowerPC this is because we now are forcing all FP constants to memory.  The
only constant for which this is interesting is zero, and only if one is storing
to memory.  XLC produces the same code as GCC mainline.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-27 19:13 -------
Another testcase which is a regression (on x86) caused by the same problem:
int g (float* g1,float t)
{
  *g1 = 0.0 + t;
}

Without -fPIC,
        fldz
        movl    4(%esp), %eax
        fadds   8(%esp)
        fstps   (%eax)
        ret
With -fPIC:
g:
        call    __i686.get_pc_thunk.cx
        addl    $_GLOBAL_OFFSET_TABLE_, %ecx
        movl    4(%esp), %eax
        flds    .LC0@GOTOFF(%ecx)
        fadds   8(%esp)
        fstps   (%eax)
        ret

in 2.95.3, we get:
g:
        fldz
        fadds 8(%esp)
        movl 4(%esp),%eax
        fstps (%eax)
        ret

But from 3.0.4 and above we get the same issue as the mainline with -fPIC.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-27 19:29 -------
Another testcase for x86 for -fPIC vs -fno-PIC:
double f(float);
double g(void)
{
  return f(0.0);
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-27 19:30 -------
(In reply to comment #4)
> Another testcase for x86 for -fPIC vs -fno-PIC:
double f(float);
double g(void)
{
  return f(1.0);
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-01 16:58 -------
Confirmed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-08-01 16:58:12
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[jakub at gcc dot gnu dot org]

------- Additional Comments From jakub at gcc dot gnu dot org  2005-09-02 15:10 -------
Testing a patch.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |jakub at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2005-08-01 16:58:12         |2005-09-02 15:10:12
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [3.4/4.0/4.1 Regression] store of 0.0 to float

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-09-06 08:53 -------
Subject: Bug 23098

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	jakub@gcc.gnu.org	2005-09-06 08:53:12

Modified files:
	gcc            : ChangeLog cse.c rtl.h simplify-rtx.c 
	gcc/testsuite  : ChangeLog 
	gcc/config/i386: i386.md 
Added files:
	gcc/testsuite/gcc.target/i386: pr23098.c 

Log message:
	PR rtl-optimization/23098
	* cse.c (fold_rtx_mem): Call delegitimize_address target hook.
	* simplify-rtx.c (constant_pool_reference_p): New function.
	* rtl.h (constant_pool_reference_p): New prototype.
	* config/i386/i386.md (pushf split, mov[sdx]f split): Use
	constant_pool_reference_p in condition and
	avoid_constant_pool_reference in preparation statements.
	
	* gcc.target/i386/pr23098.c: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.9900&r2=2.9901
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cse.c.diff?cvsroot=gcc&r1=1.359&r2=1.360
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/rtl.h.diff?cvsroot=gcc&r1=1.559&r2=1.560
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/simplify-rtx.c.diff?cvsroot=gcc&r1=1.245&r2=1.246
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.6011&r2=1.6012
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/i386/i386.md.diff?cvsroot=gcc&r1=1.653&r2=1.654
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.target/i386/pr23098.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-09-06 09:41 -------
This is now only a regression on the mainline on powerpc which still have the orginal code gen I was 
complaining about.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|powerpc-*-*, i?86-*-*       |powerpc-*-*
      Known to fail|3.3.3 3.4.0 4.0.0 4.1.0     |4.1.0
      Known to work|2.95.3 3.2.3                |2.95.3 3.2.3 3.3.3 3.4.0
                   |                            |4.0.0
            Summary|[3.4/4.0/4.1 Regression]    |[4.1 Regression] store of
                   |store of 0.0 to float       |0.0 to float


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.0.2                       |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [4.1 Regression] store of 0.0 to float

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-09-06 19:39 -------
Subject: Bug 23098

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	jakub@gcc.gnu.org	2005-09-06 19:38:59

Modified files:
	gcc/testsuite  : ChangeLog 
	gcc/testsuite/gcc.target/i386: pr23098.c 

Log message:
	PR rtl-optimization/23098
	* gcc.target/i386/pr23098.c: Add dg-require-effective-target ilp32.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.6018&r2=1.6019
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.target/i386/pr23098.c.diff?cvsroot=gcc&r1=1.1&r2=1.2



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [4.1 Regression] store of 0.0 to float

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-09-06 19:45 -------
Subject: Bug 23098

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-4_0-branch
Changes by:	jakub@gcc.gnu.org	2005-09-06 19:44:45

Modified files:
	gcc            : ChangeLog cse.c rtl.h simplify-rtx.c 
	gcc/config/i386: i386.md 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.target/i386: pr23098.c 

Log message:
	PR rtl-optimization/23098
	* cse.c (fold_rtx): Call delegitimize_address target hook.
	* simplify-rtx.c (constant_pool_reference_p): New function.
	* rtl.h (constant_pool_reference_p): New prototype.
	* config/i386/i386.md (pushf split, mov[sdx]f split): Use
	constant_pool_reference_p in condition and
	avoid_constant_pool_reference in preparation statements.
	
	* gcc.target/i386/pr23098.c: New test.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=2.7592.2.413&r2=2.7592.2.414
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/cse.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.345.2.2&r2=1.345.2.3
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/rtl.h.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.536&r2=1.536.4.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/simplify-rtx.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.230.2.7&r2=1.230.2.8
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/i386/i386.md.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.618.4.5&r2=1.618.4.6
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.5084.2.383&r2=1.5084.2.384
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.target/i386/pr23098.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=NONE&r2=1.2.2.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [4.1 Regression] store of 0.0 to float

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-09-11 23:22 -------
Subject: Bug 23098

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	dje@gcc.gnu.org	2005-09-11 23:22:11

Modified files:
	gcc            : ChangeLog 
	gcc/config/rs6000: predicates.md rs6000.md rs6000.c altivec.md 

Log message:
	PR rtl-optimization/23098
	* config/rs6000/predicates.md (easy_fp_constant): SFmode constant
	0.0f is easy.
	* config/rs6000/rs6000.md (movdf splitter): Use
	const_double_operand predicate for TARGET_POWERPC64.
	* config/rs6000/rs6000.c (rs6000_legitimize_reload_address): Expand
	SYMBOL_REF method to ABI_V4.
	
	* config/rs6000/altivec.md (build_vector_mask_for_load): Use
	replace_equiv_address.
	* config/rs6000/rs6000.c (rs6000_emit_eh_reg_restore): Mark MEM as
	rs6000_sr_alias_set and MEM_NOTRAP.
	(rs6000_aix_emit_builtin_unwind_init): Use gen_frame_mem().

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.9933&r2=2.9934
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/rs6000/predicates.md.diff?cvsroot=gcc&r1=1.25&r2=1.26
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/rs6000/rs6000.md.diff?cvsroot=gcc&r1=1.406&r2=1.407
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/rs6000/rs6000.c.diff?cvsroot=gcc&r1=1.866&r2=1.867
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/rs6000/altivec.md.diff?cvsroot=gcc&r1=1.44&r2=1.45



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [4.1 Regression] store of 0.0 to float

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-09-12 15:12 -------
Fixed.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098

[Bug rtl-optimization/23098] [4.1 Regression] store of 0.0 to float

[sayle at gcc dot gnu dot org]

------- Comment #14 from sayle at gcc dot gnu dot org  2005-12-31 00:39 -------
Subject: Bug 23098

Author: sayle
Date: Sat Dec 31 00:39:42 2005
New Revision: 109192

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=109192
Log:

        PR target/25213
        Backport from mainline
        2005-09-06  Jakub Jelinek  <jakub@redhat.com>

        PR rtl-optimization/23098
        * cse.c (fold_rtx_mem): Call delegitimize_address target hook.
        * simplify-rtx.c (constant_pool_reference_p): New function.
        * rtl.h (constant_pool_reference_p): New prototype.
        * config/i386/i386.md (pushf split, mov[sdx]f split): Use
        constant_pool_reference_p in condition and
        avoid_constant_pool_reference in preparation statements.

        * gcc.dg/pr23098.c: Backport testcase from mainline.


Added:
    branches/gcc-3_4-branch/gcc/testsuite/gcc.dg/pr23098.c
Modified:
    branches/gcc-3_4-branch/gcc/ChangeLog
    branches/gcc-3_4-branch/gcc/config/i386/i386.md
    branches/gcc-3_4-branch/gcc/cse.c
    branches/gcc-3_4-branch/gcc/rtl.h
    branches/gcc-3_4-branch/gcc/simplify-rtx.c
    branches/gcc-3_4-branch/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23098