[Bug tree-optimization/30930] New: [4.3 Regression] vector can cause to create an extra variable

[Bug tree-optimization/30930] New: [4.3 Regression] vector can cause to create an extra variable

[pinskia at gcc dot gnu dot org]

While trying to make DECL_GIMPLE_REG_P more generic, I ran into a regression
due to we never change non gimple register variables into gimple registers so I
looked into it and found that I could also reproduce it in the current sources
with vector types (complex does not matter as much as they are always SRA'd).
Testcase:
#define vector __attribute__(( vector_size(16) ))

vector float f(vector float a, int b, vector float c)
{
  vector float dd = c*a;
  if (0)
  {
   vector float *d = &a;
   vector float *d1 = &c;
    g:;
  }
  if (b)
    return dd * a;
  return c*a;
}


You will see that a is never turned into a gimple register which could cause
some issues with debugging info and also this introduces an extra variable
which could cause a memory usage/compile time slow down.


-- 
           Summary: [4.3 Regression] vector can cause to create an extra
                    variable
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pinskia at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.3.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2007-02-23 09:56 -------
I believe the mini-pre-alias passes from honza will fix this by recomputing
addressability.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2007-02-23 22:47 -------
(In reply to comment #1)
> I believe the mini-pre-alias passes from honza will fix this by recomputing
> addressability.

But this is not about recomputing addressability but instead recomputing
DECL_GIMPLE_REG_P which we don't do currently.  We could do it the same time as
we recompute addressability and I think we should.  I will go look into this
futher later today.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable

[mmitchel at gcc dot gnu dot org]

-- 

mmitchel at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2008-01-13 14:46 -------
It indeed works on the 4.2 branch where the loads magically disappear with the
first may_alias pass.  Possibly because it doesn't pay attention to whether
DECL_GIMPLE_REG is set or not.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
           Keywords|                            |missed-optimization
      Known to work|                            |4.2.3
   Last reconfirmed|0000-00-00 00:00:00         |2008-01-13 14:46:12
               date|                            |
            Summary|[4.3 Regression] vector can |[4.3 Regression] vector can
                   |cause to create an extra    |cause to create an extra
                   |variable                    |variable, DECL_GIMPLE_REG_P
                   |                            |not recomputed


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2008-01-13 17:57 -------
(In reply to comment #3)
> It indeed works on the 4.2 branch where the loads magically disappear with the
> first may_alias pass.  Possibly because it doesn't pay attention to whether
> DECL_GIMPLE_REG is set or not.

Well DECL_GIMPLE_REG was not needed for vector types before 4.3.0 :) so yes it
was ignored.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[jsm28 at gcc dot gnu dot org]

------- Comment #5 from jsm28 at gcc dot gnu dot org  2008-03-15 00:40 -------
Update milestone after 4.3.0 release.


-- 

jsm28 at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.3.0                       |4.3.1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[rguenth at gcc dot gnu dot org]

------- Comment #6 from rguenth at gcc dot gnu dot org  2008-06-06 14:56 -------
4.3.1 is being released, adjusting target milestone.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.3.1                       |4.3.2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[jsm28 at gcc dot gnu dot org]

------- Comment #7 from jsm28 at gcc dot gnu dot org  2008-08-27 22:01 -------
4.3.2 is released, changing milestones to 4.3.3.


-- 

jsm28 at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.3.2                       |4.3.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #8 from pinskia at gcc dot gnu dot org  2008-09-17 22:28 -------
Created an attachment (id=16349)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=16349&action=view)
Patch which piggy backs on honza's addressable pass

This is a patch which uses honza's mini-addressable pass to the same thing for
DECL_GIMPLE_REG_P.  It can be used for the generic case instead of just vector
types which will help out the VCE on the LHS side issue.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #9 from pinskia at gcc dot gnu dot org  2008-09-17 22:40 -------
Note I needed to add a check to make sure we are only doing this currently for
complex and vector types since those are the ones that have DECL_GIMPLE_REG_P
currently.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #10 from pinskia at gcc dot gnu dot org  2008-09-17 22:50 -------
Here is a testcase where we were missing an optimization at the tree level
because of DECL_GIMPLE_REG_P:
#define vector __attribute__(( vector_size(16) ))

float f(vector float a, int b, vector float c)
{
  vector float dd = c*a;
  a = (vector float){0,0,0,0};
  c = (vector float){0,0,0,0};
  {
   float d = ((float*)&a)[0];
   float d1 = ((float*)&c)[0];
   return d*d1;
  }
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #11 from pinskia at gcc dot gnu dot org  2008-09-18 01:00 -------
Here is a testcase which shows a missed optimization in general (on both the
tree level and the RTL one) for x86 with SSE2:
#define vector __attribute((vector_size(16) ))
vector float a;

float f(float b)
{
  vector float c = {0, 0, 0, 0};
  vector float d = {0, 0, 0, 0};
  d += c;
  return ((float*)&c)[2];
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3/4.4 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[rguenther at suse dot de]

------- Comment #12 from rguenther at suse dot de  2008-09-18 08:23 -------
Subject: Re:  [4.3/4.4 Regression] vector can
 cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

On Wed, 17 Sep 2008, pinskia at gcc dot gnu dot org wrote:

> ------- Comment #10 from pinskia at gcc dot gnu dot org  2008-09-17 22:50 -------
> Here is a testcase where we were missing an optimization at the tree level
> because of DECL_GIMPLE_REG_P:
> #define vector __attribute__(( vector_size(16) ))
> 
> float f(vector float a, int b, vector float c)
> {
>   vector float dd = c*a;
>   a = (vector float){0,0,0,0};
>   c = (vector float){0,0,0,0};
>   {
>    float d = ((float*)&a)[0];
>    float d1 = ((float*)&c)[0];
>    return d*d1;
>   }
> }

I wonder why we not just can add both, TREE_ADDRESSABLE and
DECL_GIMPLE_REG_P ...

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #13 from pinskia at gcc dot gnu dot org  2008-09-19 22:24 -------
Fixed on the trunk.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|4.2.3                       |4.2.3 4.4.0
            Summary|[4.3/4.4 Regression] vector |[4.3 Regression] vector can
                   |can cause to create an extra|cause to create an extra
                   |variable, DECL_GIMPLE_REG_P |variable, DECL_GIMPLE_REG_P
                   |not recomputed              |not recomputed


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #14 from pinskia at gcc dot gnu dot org  2008-09-19 22:25 -------
Subject: Bug 30930

Author: pinskia
Date: Fri Sep 19 22:24:06 2008
New Revision: 140501

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=140501
Log:
2008-09-19  Andrew Pinski  <andrew_pinski@playstation.sony.com>

        PR tree-opt/30930
        * tree-ssa.c (execute_update_addresses_taken): Also update
        DECL_GIMPLE_REG_P for vector and complex types.


2008-09-19  Andrew Pinski  <andrew_pinski@playstation.sony.com>

        PR tree-opt/30930
        * gcc.dg/tree-ssa/vector-2.c: New test.
        * gcc.dg/tree-ssa/vector-3.c: New test.


Added:
    trunk/gcc/testsuite/gcc.dg/tree-ssa/vector-2.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/vector-3.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-ssa.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[rguenth at gcc dot gnu dot org]

------- Comment #15 from rguenth at gcc dot gnu dot org  2009-01-24 10:19 -------
GCC 4.3.3 is being released, adjusting target milestone.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.3.3                       |4.3.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[pinskia at gcc dot gnu dot org]

------- Comment #16 from pinskia at gcc dot gnu dot org  2009-04-16 15:42 -------
I am no longer working on this one.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |4.3.3
           Priority|P4                          |P2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[rguenth at gcc dot gnu dot org]

------- Comment #17 from rguenth at gcc dot gnu dot org  2009-08-04 12:28 -------
GCC 4.3.4 is being released, adjusting target milestone.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.3.4                       |4.3.5


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930

[Bug tree-optimization/30930] [4.3 Regression] vector can cause to create an extra variable, DECL_GIMPLE_REG_P not recomputed

[rguenth at gcc dot gnu dot org]

------- Comment #18 from rguenth at gcc dot gnu dot org  2010-04-20 13:27 -------
WONTFIX on the 4.3 branch.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
      Known to fail|4.3.3                       |4.3.3 4.3.4
         Resolution|                            |FIXED
   Target Milestone|4.3.5                       |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30930