[Bug middle-end/28075] New: gimplifier introduces unnecessary type conversions

[Bug middle-end/28075] New: gimplifier introduces unnecessary type conversions

[rguenth at gcc dot gnu dot org]

typedef struct {
  double min;
  double max;
} interval;
inline interval add(interval x, interval y)
{
  interval r;
  r.min = x.min + y.min;
  r.max = x.max + y.max;
  return r;
}
interval foo (interval a, interval b, interval c)
{
  return add (a, add (b, c));
}

for the temporary for gimplifying the nested call to add, the gimplifier
introduces a temporary of type TYPE_MAIN_VARIANT (interval), which survives
as added casts.

This happens here:

static inline tree
create_tmp_from_val (tree val)
{
  return create_tmp_var (TYPE_MAIN_VARIANT (TREE_TYPE (val)), get_name (val));
}

in the optimized dump we still have

foo (a, b, c)
{
  double r$min;
  double r$min;
  double a$max;
  double a$min;
  struct interval y;
  struct
  {
    double min;
    double max;
  } D.1542;

<bb 2>:
  a$max = a.max;
  a$min = a.min;
  r$min = b.min + c.min;
  D.1542.max = b.max + c.max;
  D.1542.min = r$min;
  y = (struct interval) D.1542;
  r$min = y.min + a$min;
  <retval>.max = y.max + a$max;
  <retval>.min = r$min;
  return <retval>;

}

by simply removing this TYPE_MAIN_VARAINT there we end up with two
temporary structures less:

Analyzing Edge Insertions.
foo (a, b, c)
{
  double r$min;

<bb 2>:
  r$min = b.min + c.min + a.min;
  <retval>.max = b.max + c.max + a.max;
  <retval>.min = r$min;
  return <retval>;

}


-- 
           Summary: gimplifier introduces unnecessary type conversions
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Keywords: missed-optimization, memory-hog, compile-time-hog
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: rguenth at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] gimplifier introduces unnecessary type conversions

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2006-06-17 19:44 -------
http://gcc.gnu.org/ml/gcc-patches/2006-05/msg01236.html does not help (we don't
FRE structure assignments).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] gimplifier introduces unnecessary type conversions

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2006-06-17 19:45 -------
I have a patch to fix the inliner to get rid of the unnecessary type
conversions.

This only happens with C front-end.


-- 

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|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2006-06-17 19:45:38
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] gimplifier introduces unnecessary type conversions

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2006-06-17 19:51 -------
This sort of blocks 28071 because we cannot optimize anything with the casts
there and so regalloc blows up with the large BB.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|                            |28071
              nThis|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] inliner introduces unnecessary type conversions

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2006-06-17 19:53 -------
Here is the patch:
Index: tree-inline.c
===================================================================
--- tree-inline.c       (revision 114740)
+++ tree-inline.c       (working copy)
@@ -1081,6 +1081,8 @@ setup_one_parameter (copy_body_data *id,

       if (rhs == error_mark_node)
        return;
+
+      STRIP_USELESS_TYPE_CONVERSION (rhs);

       /* We want to use MODIFY_EXPR, not INIT_EXPR here so that we
         keep our trees in gimple form.  */
@@ -1267,6 +1269,8 @@ declare_return_variable (copy_body_data 
   use = var;
   if (!lang_hooks.types_compatible_p (TREE_TYPE (var), caller_type))
     use = fold_convert (caller_type, var);
+    
+  STRIP_USELESS_TYPE_CONVERSION (use);

  done:
   /* Register the VAR_DECL as the equivalent for the RESULT_DECL; that


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
OtherBugsDependingO|28071                       |
              nThis|                            |
            Summary|gimplifier introduces       |inliner introduces
                   |unnecessary type conversions|unnecessary type conversions


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1/4.2 Regression] inliner introduces unnecessary type conversions

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|inliner introduces          |[4.1/4.2 Regression] inliner
                   |unnecessary type conversions|introduces unnecessary type
                   |                            |conversions
   Target Milestone|---                         |4.1.2


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1/4.2 Regression] inliner introduces unnecessary type conversions

[pinskia at gcc dot gnu dot org]

------- Comment #5 from pinskia at gcc dot gnu dot org  2006-06-20 02:10 -------
Subject: Bug 28075

Author: pinskia
Date: Tue Jun 20 02:09:57 2006
New Revision: 114801

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=114801
Log:
2006-06-19  Andrew Pinski  <pinskia@gmail.com>

        PR middle-end/28075
        * tree-inline.c (setup_one_parameter): Strip useless
        type conversion before adding it to the IR.
        (declare_return_variable): Likewise.


2006-06-19  Andrew Pinski  <pinskia@gmail.com>

        PR middle-end/28075
        * gcc.dg/tree-ssa/inline-1.c: New test.



Added:
    trunk/gcc/testsuite/gcc.dg/tree-ssa/inline-1.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-inline.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[pinskia at gcc dot gnu dot org]

------- Comment #6 from pinskia at gcc dot gnu dot org  2006-06-20 02:11 -------
Fixed on the mainline and will commit on the 4.1 after two weeks (unless I get
some time during the summit to commit it, I will).


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |4.1.1
      Known to work|                            |4.0.2 4.2.0
            Summary|[4.1/4.2 Regression] inliner|[4.1 Regression] inliner
                   |introduces unnecessary type |introduces unnecessary type
                   |conversions                 |conversions


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[danglin at gcc dot gnu dot org]

------- Comment #7 from danglin at gcc dot gnu dot org  2006-06-26 13:45 -------
This test is still failing on hppa64-hp-hpux11.11.


-- 

danglin at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |danglin at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[dave at hiauly1 dot hia dot nrc dot ca]

------- Comment #8 from dave at hiauly1 dot hia dot nrc dot ca  2006-06-26 13:47 -------
Subject: Re:  [4.1 Regression] inliner introduces unnecessary type conversions

Attached dump.

Dave


------- Comment #9 from dave at hiauly1 dot hia dot nrc dot ca  2006-06-26 13:47 -------
Created an attachment (id=11757)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=11757&action=view)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[pinskia at gmail dot com]

------- Comment #10 from pinskia at gmail dot com  2006-06-26 16:02 -------
Subject: Re:  [4.1 Regression] inliner introduces unnecessary type conversions


On Jun 26, 2006, at 6:45 AM, danglin at gcc dot gnu dot org wrote:

>
>
> ------- Comment #7 from danglin at gcc dot gnu dot org  2006-06-26  
> 13:45 -------
> This test is still failing on hppa64-hp-hpux11.11.

Yes the test has a simple problem of also matching the type even  
without the "()",
I have a fix but I have not applied it yet.

-- Pinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[patchapp at dberlin dot org]

------- Comment #11 from patchapp at dberlin dot org  2006-06-29 21:36 -------
Subject: Bug number PR middle-end/28075

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2006-06/msg00964.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[pinskia at gcc dot gnu dot org]

------- Comment #12 from pinskia at gcc dot gnu dot org  2006-06-30 14:31 -------
(In reply to comment #10)
> I have a fix but I have not applied it yet.
And I just applied it.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[hubicka at gcc dot gnu dot org]

------- Comment #13 from hubicka at gcc dot gnu dot org  2006-07-22 12:42 -------
Is this bug to be closed then?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[rguenth at gcc dot gnu dot org]

------- Comment #14 from rguenth at gcc dot gnu dot org  2006-07-22 18:53 -------
A fix for the failing testcase was applied, the bug is still open on the 4.1
branch.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[pinskia at gcc dot gnu dot org]

------- Comment #15 from pinskia at gcc dot gnu dot org  2006-08-14 14:13 -------
Subject: Bug 28075

Author: pinskia
Date: Mon Aug 14 15:13:29 2006
New Revision: 116127

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=116127
Log:
+2006-08-14  Andrew Pinski  <pinskia@gmail.com>
+
+       * gcc.dg/tree-ssa/inline-1.c: Fix up scan-tree-dump-times
+       for '(' / ')' needing an extra '\'.
+
+2006-08-14  Andrew Pinski  <pinskia@gmail.com>
+
+       PR middle-end/28075
+       * gcc.dg/tree-ssa/inline-1.

+2006-08-14  Andrew Pinski  <pinskia@gmail.com>
+
+       PR middle-end/28075
+       * tree-inline.c (setup_one_parameter): Strip useless
+       type conversion before adding it to the IR.
+       (declare_return_variable): Likewise.
+


Added:
    branches/gcc-4_1-branch/gcc/testsuite/gcc.dg/tree-ssa/inline-1.c
      - copied, changed from r114801,
trunk/gcc/testsuite/gcc.dg/tree-ssa/inline-1.c
Modified:
    branches/gcc-4_1-branch/gcc/ChangeLog
    branches/gcc-4_1-branch/gcc/testsuite/ChangeLog
    branches/gcc-4_1-branch/gcc/tree-inline.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[pinskia at gcc dot gnu dot org]

------- Comment #16 from pinskia at gcc dot gnu dot org  2006-08-14 14:15 -------
Fixed.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075

[Bug middle-end/28075] [4.1 Regression] inliner introduces unnecessary type conversions

[ebotcazou at gcc dot gnu dot org]

------- Comment #17 from ebotcazou at gcc dot gnu dot org  2006-08-26 14:00 -------
Andrew, please remove the superfluous TAB you've added in 

+       * tree-inline.c (setup_one_parameter): Strip useless
+       type conversion before adding it to the IR.

on the 4.1 branch at least.  Thanks in advance.


-- 

ebotcazou at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ebotcazou at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28075