[Bug rtl-optimization/34012] New: [4.3 Regression] Pessimization caused by fwprop

[Bug rtl-optimization/34012] New: [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

On
void bar (long int *);
void foo (void)
{
  long int buf[10];
  __builtin_memset (buf, 8, sizeof (buf));
  bar (buf);
}
fwprop effectively undoes what CSE optimized.  0x0808080808080808 is not a
valid constant for movdi_1_rex64 when the target is memory (movdi_1_rex64 only
allows non-32-bit-sign-extended immediates in source if the destination is a
register).


-- 
           Summary: [4.3 Regression] Pessimization caused by fwprop
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: rtl-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jakub at gcc dot gnu dot org
GCC target triplet: x86_64-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

-- 

jakub at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2007-11-07 10:17:34
               date|                            |
   Target Milestone|---                         |4.3.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

------- Comment #1 from jakub at gcc dot gnu dot org  2007-11-07 10:24 -------
With 4.2 and earlier, we get:
        movabsq $578721382704613384, %rax
        movq    %rsp, %rdi
        movq    %rax, (%rsp)
        movq    %rax, 8(%rsp)
        movq    %rax, 16(%rsp)
        movq    %rax, 24(%rsp)
        movq    %rax, 32(%rsp)
        movq    %rax, 40(%rsp)
        movq    %rax, 48(%rsp)
        movq    %rax, 56(%rsp)
        movq    %rax, 64(%rsp)
        movq    %rax, 72(%rsp)
        call    bar
with the trunk:
        movabsq $578721382704613384, %rax
        movabsq $578721382704613384, %r11
        movq    %rax, (%rsp)
        movabsq $578721382704613384, %rax
        movq    %rsp, %rdi
        movq    %rax, 8(%rsp)
        movabsq $578721382704613384, %r10
        movabsq $578721382704613384, %r9
        movabsq $578721382704613384, %r8
        movabsq $578721382704613384, %rsi
        movabsq $578721382704613384, %rcx
        movabsq $578721382704613384, %rdx
        movabsq $578721382704613384, %rax
        movq    %r11, 16(%rsp)
        movq    %r10, 24(%rsp)
        movq    %r9, 32(%rsp)
        movq    %r8, 40(%rsp)
        movq    %rsi, 48(%rsp)
        movq    %rcx, 56(%rsp)
        movq    %rdx, 64(%rsp)
        movq    %rax, 72(%rsp)
        call    bar
which is far worse.  The predicate on the source operand on movdi_1_rex64
allows it (general_operand), but the constraints don't (but letting forwprop to
try all constraints to see what is and is not desirable would be weird). 
Alternatively, perhaps forwprop could try to emit_move_insn and see if it
doesn't create more insns than the one it is substing into, if more than one,
it should just punt, otherwise it can just subst into it.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[bonzini at gnu dot org]

------- Comment #2 from bonzini at gnu dot org  2007-11-07 11:03 -------
fwprop should check costs just like combine does.  Unfortunately the cost do
need a little bit of tweaking even if one implements the idea.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[bonzini at gnu dot org]

------- Comment #3 from bonzini at gnu dot org  2007-11-07 11:04 -------
Created an attachment (id=14495)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14495&action=view)
patch that fixes the bug


I'm not sure about the correctness of the i386.c hunk.  The problem is that a
cost of 0 is mapped by insn_rtx_cost to COSTS_N_INSNS(1) == 4.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

------- Comment #4 from jakub at gcc dot gnu dot org  2007-11-07 13:40 -------
BTW, why don't you use just rtx_cost instead of insn_rtx_cost?
In each case you have an insn, so you can do single_set on it and run
rtx_cost (SET_SRC (set), SET) on it directly.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[paolo dot bonzini at lu dot unisi dot ch]

------- Comment #5 from paolo dot bonzini at lu dot unisi dot ch  2007-11-07 13:53 -------
Subject: Re:  [4.3 Regression] Pessimization caused
 by fwprop

> BTW, why don't you use just rtx_cost instead of insn_rtx_cost?
> In each case you have an insn, so you can do single_set on it and run
> rtx_cost (SET_SRC (set), SET) on it directly.

You're right.  I was just mimicking combine to see if the problem was 
simple to solve or more deeply rooted.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[bonzini at gnu dot org]

------- Comment #6 from bonzini at gnu dot org  2007-11-07 17:02 -------
Created an attachment (id=14499)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14499&action=view)
patch using rtx_cost

This patch uses rtx_cost instead of insn_rtx_cost, which is faster and does not
require touching i386.c.


-- 

bonzini at gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #14495|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

------- Comment #7 from jakub at gcc dot gnu dot org  2007-11-07 20:18 -------
Created an attachment (id=14502)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14502&action=view)
gcc43-pr34012.patch

Updated patch with testcase.  Paolo, are you bootstrapping/regtesting this or
should I?  I can do that on x86_64-linux, ppc64-linux and ia64-linux overnight.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[paolo dot bonzini at lu dot unisi dot ch]

------- Comment #8 from paolo dot bonzini at lu dot unisi dot ch  2007-11-08 06:10 -------
Subject: Re:  [4.3 Regression] Pessimization caused
 by fwprop

jakub at gcc dot gnu dot org wrote:
> ------- Comment #7 from jakub at gcc dot gnu dot org  2007-11-07 20:18 -------
> Created an attachment (id=14502)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14502&action=view)
>  --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14502&action=view)
> gcc43-pr34012.patch
> 
> Updated patch with testcase.  Paolo, are you bootstrapping/regtesting this or
> should I?  I can do that on x86_64-linux, ppc64-linux and ia64-linux overnight.

I could have done it only on Monday and only for x86_64 (CPU time for 
the local x86_64 is "reserved" by another guy till then...) so yes, it's 
much better if you do it.  Thanks!

Paolo


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

------- Comment #9 from jakub at gcc dot gnu dot org  2007-11-08 17:05 -------
Bootstrapped/regtested on x86_64-linux, ppc64-linux (-m64 default) and
ia64-linux.  No regressions.  Could you please send it to gcc-patches?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

------- Comment #10 from jakub at gcc dot gnu dot org  2007-11-09 13:03 -------
Subject: Bug 34012

Author: jakub
Date: Fri Nov  9 13:02:25 2007
New Revision: 130043

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=130043
Log:
        PR rtl-optimization/34012
        * fwprop.c (try_fwprop_subst): Do not replace if the new
        SET_SRC has a higher cost than the old one.

        * gcc.target/i386/pr34012.c: New test.

Added:
    trunk/gcc/testsuite/gcc.target/i386/pr34012.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/fwprop.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012

[Bug rtl-optimization/34012] [4.3 Regression] Pessimization caused by fwprop

[jakub at gcc dot gnu dot org]

------- Comment #11 from jakub at gcc dot gnu dot org  2007-11-09 13:05 -------
Fixed.


-- 

jakub at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34012