[Bug ada/45830] New: [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[Bug ada/45830] New: [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

           Summary: [4.4/4.5/4.6 Regression] Code+rodata increase with
                    -ftree-switch-conversion
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ada
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: jakub@gcc.gnu.org
                CC: jamborm@gcc.gnu.org


int
foo (int *a)
{
  switch (*a)
    {
    case 0:
    case 3:
    case 1:
    case 2:
    case 4:
    case 23:
    case 26:
    case 19:
    case 5:
    case 21:
    case 20:
    case 22:
    case 27:
      return 1;
    default:
      return 0;
    }
}

generates significantly shorter code+rodata with -fno-tree-switch-conversion
than without.  I guess for the cases where 
      if (CASE_USE_BIT_TESTS
          && ! TREE_CONSTANT (index_expr)
          && compare_tree_int (range, GET_MODE_BITSIZE (word_mode)) < 0
          && compare_tree_int (range, 0) > 0
          && lshift_cheap_p ()
          && ((uniq == 1 && count >= 3)
              || (uniq == 2 && count >= 5)
              || (uniq == 3 && count >= 6)))
is true tree-switch-conversion.c should not perform the optimization, and
in any case the question is if the table needs to be created always for the
type of the constants, even when they are integral and would fit into narrower
integer type (like char/unsigned char in this case).

[Bug tree-optimization/45830] [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2010.09.29 13:30:53
               date|                            |
          Component|ada                         |tree-optimization
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2010-09-29 13:30:53 UTC ---
Confirmed.

I think switch-conversion should be merged with whatever tree-level
switch-lowering/expansion code we end up with.

[Bug tree-optimization/45830] [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.4.5

[Bug tree-optimization/45830] [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug tree-optimization/45830] [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.4.5                       |4.4.6

[Bug tree-optimization/45830] [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-19 16:17:06 UTC ---
Created attachment 22459
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=22459
gcc46-pr45830.patch

Untested fix that should cure both reported issues, i.e. don't optimize and let
switch being expanded as bit test if beneficial, and use smaller integral types
when possible for the CSWTCH arrays.

[Bug tree-optimization/45830] [4.4/4.5/4.6 Regression] Code+rodata increase with -ftree-switch-conversion

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-19 23:49:03 UTC ---
Author: jakub
Date: Fri Nov 19 23:48:57 2010
New Revision: 166966

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=166966
Log:
    PR tree-optimization/45830
    * stmt.c (expand_switch_using_bit_tests_p): New function.
    (expand_case): Use it.
    * tree.h (expand_switch_using_bit_tests_p): New prototype.
    * tree-switch-conversion.c (struct switch_conv_info): Add
    bit_test_uniq, bit_test_count and bit_test_bb fields.
    (check_range): Fix a comment.
    (check_process_case): Compute bit_test_uniq and bit_test_count.
    (create_temp_arrays): Use XCNEWVEC, merge 3 arrays into one
    allocation.
    (free_temp_arrays): Use XDELETEVEC, adjust for the 3 arrays merging.
    (constructor_contains_same_values_p): Use FOR_EACH_VEC_ELT.
    (array_value_type): New function.
    (build_one_array): Use it, if it returned different type,
    fold_convert all constructor fields and convert back to the
    wider type in the generated code.
    (process_switch): Initialize bit_test_uniq, bit_test_count and
    bit_test_bb fields.  Don't optimize if expand_switch_using_bit_tests_p
    returned true.

    * gcc.target/i386/pr45830.c: New test.
    * gcc.c-torture/execute/pr45830.c: New test.

Added:
    trunk/gcc/testsuite/gcc.c-torture/execute/pr45830.c
    trunk/gcc/testsuite/gcc.target/i386/pr45830.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/stmt.c
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-switch-conversion.c
    trunk/gcc/tree.h

[Bug tree-optimization/45830] [4.4/4.5 Regression] Code+rodata increase with -ftree-switch-conversion

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.4/4.5/4.6 Regression]    |[4.4/4.5 Regression]
                   |Code+rodata increase with   |Code+rodata increase with
                   |-ftree-switch-conversion    |-ftree-switch-conversion

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2010-11-19 23:53:04 UTC ---
Fixed on the trunk.

[Bug tree-optimization/45830] [4.4/4.5 Regression] Code+rodata increase with -ftree-switch-conversion

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.4.6                       |4.4.7

[Bug tree-optimization/45830] [4.4/4.5 Regression] Code+rodata increase with -ftree-switch-conversion

[gjl at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

--- Comment #5 from Georg-Johann Lay <gjl at gcc dot gnu.org> 2011-12-12 20:13:45 UTC ---
Author: gjl
Date: Mon Dec 12 20:13:40 2011
New Revision: 182262

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=182262
Log:
    PR tree-optimization/45830
    * gcc.c-torture/execute/pr45830.c: Move from here to...
    * gcc.dg/torture/pr45830.c: ...this new file.  Add dg-do and
    dg-additional-options magic.


Added:
    trunk/gcc/testsuite/gcc.dg/torture/pr45830.c
      - copied, changed from r182043,
trunk/gcc/testsuite/gcc.c-torture/execute/pr45830.c
Removed:
    trunk/gcc/testsuite/gcc.c-torture/execute/pr45830.c
Modified:
    trunk/gcc/testsuite/ChangeLog

[Bug tree-optimization/45830] [4.5 Regression] Code+rodata increase with -ftree-switch-conversion

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.4.7                       |4.5.4

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-03-13 12:46:58 UTC ---
4.4 branch is being closed, moving to 4.5.4 target.

[Bug tree-optimization/45830] [4.5 Regression] Code+rodata increase with -ftree-switch-conversion

[steven at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Steven Bosscher <steven at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |steven at gcc dot gnu.org

--- Comment #7 from Steven Bosscher <steven at gcc dot gnu.org> 2012-04-18 20:33:05 UTC ---
*** Bug 53034 has been marked as a duplicate of this bug. ***

[Bug tree-optimization/45830] [4.5 Regression] Code+rodata increase with -ftree-switch-conversion

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45830

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|4.5.4                       |4.6.0

--- Comment #8 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-07-02 10:39:49 UTC ---
Fixed in 4.6.0, the 4.5 branch is being closed.