[Bug tree-optimization/18589] could optimize FP multiplies better

[Bug tree-optimization/18589] could optimize FP multiplies better

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2005-12-21 03:49:53         |2012-01-04

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-01-05 23:06:56 UTC ---
The tree level on the trunk we get:
  powmult.2_10 = foo_1(D) * foo_1(D);
  D.1709_4 = bar_3(D) * bar_3(D);
  D.1710_5 = D.1709_4 * bar_3(D);
  D.1711_6 = D.1710_5 * bar_3(D);
  powmult.2_9 = D.1711_6 * powmult.2_10;
  D.1707_7 = powmult.2_9 * powmult.2_10;
--- CUT ---
which is just as bad.

[Bug tree-optimization/18589] could optimize FP multiplies better

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu.org,
                   |                            |wschmidt at gcc dot gnu.org

--- Comment #6 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-01-09 12:20:31 UTC ---
Yeah, reassoc does not canonicalize to pow () so the tree level optimal
expansion does not trigger [in reality reassoc should probably do both
on-the-fly - linearly expand existing pow()s to expose them to multiply
chains and sorting, combine them back and then emit them in optimal form].

Bill, something you want to tackle?

[Bug tree-optimization/18589] could optimize FP multiplies better

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

--- Comment #7 from William J. Schmidt <wschmidt at gcc dot gnu.org> 2012-01-09 13:06:34 UTC ---
Sure, I'll at least have a look at it when I get some time.

[Bug tree-optimization/18589] could optimize FP multiplies better

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

William J. Schmidt <wschmidt at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |wschmidt at gcc dot gnu.org
                   |gnu.org                     |
   Target Milestone|---                         |4.8.0

--- Comment #8 from William J. Schmidt <wschmidt at gcc dot gnu.org> 2012-01-13 22:27:48 UTC ---
I've started to look at this -- I'll plan to get a patch in place for 4.8.

[Bug tree-optimization/18589] could optimize FP multiplies better

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

--- Comment #9 from William J. Schmidt <wschmidt at gcc dot gnu.org> 2012-04-12 16:15:24 UTC ---
Author: wschmidt
Date: Thu Apr 12 16:15:13 2012
New Revision: 186384

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=186384
Log:
gcc:

2012-04-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

    PR tree-optimization/18589
    * tree-ssa-reassoc.c (reassociate_stats): Add two fields.
    (operand_entry): Add count field.
    (add_repeat_to_ops_vec): New function.
    (completely_remove_stmt): Likewise.
    (remove_def_if_absorbed_call): Likewise.
    (remove_visited_stmt_chain): Remove feeding builtin pow/powi calls.
    (acceptable_pow_call): New function.
    (linearize_expr_tree): Look for builtin pow/powi calls and add operand
    entries with repeat counts when found.
    (repeat_factor_d): New struct and associated typedefs.
    (repeat_factor_vec): New static vector variable.
    (compare_repeat_factors): New function.
    (get_reassoc_pow_ssa_name): Likewise.
    (attempt_builtin_powi): Likewise.
    (reassociate_bb): Call attempt_builtin_powi.
    (fini_reassoc): Two new calls to statistics_counter_event.

gcc/testsuite:

2012-04-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>

    PR tree-optimization/18589
    * gcc.dg/tree-ssa/pr18589-1.c: New test.
    * gcc.dg/tree-ssa/pr18589-2.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-3.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-4.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-5.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-6.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-7.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-8.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-9.c: Likewise.
    * gcc.dg/tree-ssa/pr18589-10.c: Likewise.



Added:
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-1.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-10.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-2.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-3.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-4.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-5.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-6.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-7.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-8.c
    trunk/gcc/testsuite/gcc.dg/tree-ssa/pr18589-9.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-ssa-reassoc.c

[Bug tree-optimization/18589] could optimize FP multiplies better

[wschmidt at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

William J. Schmidt <wschmidt at gcc dot gnu.org> changed:

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

--- Comment #10 from William J. Schmidt <wschmidt at gcc dot gnu.org> 2012-04-12 16:16:50 UTC ---
Fixed.

[Bug tree-optimization/18589] could optimize FP multiplies better

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2006-03-05 17:43 -------
In 4.2.0 and above we get:
baz:
.LFB2:
        mulsd   %xmm1, %xmm0
        mulsd   %xmm0, %xmm0
        mulsd   %xmm0, %xmm0
        ret

Which is what I recommend but we don't get that on the tree level:
  return bar * bar * __builtin_pow (foo, 4.0e+0) * bar * bar;

So this is just a tree level missed optimization.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |TREE


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

[Bug tree-optimization/18589] could optimize FP multiplies better

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-05 19:37 -------
I think PR 22312 mentions what the current problem with reassoc is (well once I submit the patch to 
introduce reassociation for fp).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |22312


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589

[Bug tree-optimization/18589] could optimize FP multiplies better

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-01-12 06:50 -------
This is actually not a target issue, it can be shown on ppc also and other targets including x86.
doing (f1*f2)^2^2 will be the best every where as it is only three instructions and it would take the 
same time as what is proposed if there are two FPU units  but what I said is the smallest and fastest 
version.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|target                      |tree-optimization
  GCC build triplet|amd64-linux                 |
   GCC host triplet|amd64-linux                 |
 GCC target triplet|amd64-linux                 |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18589