[Bug tree-optimization/38721] [alias-improvements] vectorizer miscompiles gfortran.fortran-torture/execute/elemental.f90 at -O3

2009-01-05 Thread irar at il dot ibm dot com 


--- Comment #1 from irar at il dot ibm dot com  2009-01-05 13:19 ---
Here is a reduced testcase:

program test_elemental
   implicit none
   integer, dimension (2, 4) :: a
   integer, dimension (2, 4) :: b
   integer(kind = 8), dimension(2) :: c

   a = reshape ((/2, 3, 4, 5, 6, 7, 8, 9/), (/2, 4/))
   b = 0
   a = e_fn (a(:, 4:1:-1), 1 + b)
   ! This tests intrinsic elemental conversion functions.
   c = 2 * a(1, 1)
   if (any (c .ne. 14)) call abort

   ! This triggered bug due to building ss chains in the wrong order.
   b = 0;
   a = a - e_fn (a, b)
   if (any (a .ne. 0)) call abort

contains

elemental integer(kind=4) function e_fn (p, q)
   integer, intent(in) :: p, q
   e_fn = p - q
end function
end program


The problem is that dse2 removes the stores to array A.4 which is used by the
vectorized code:

  A.4[0] = D.1635_155;
  ...
  A.4[7] = D.1635_165;
  vect_pA.67_156 = (vector integer(kind=4) *) &A.4;
  vect_pa.73_197 = (vector integer(kind=4) *) &a;
  vect_var_.68_254 = *vect_pA.67_156;
  *vect_pa.73_197 = vect_var_.68_254;
  vect_pA.63_256 = vect_pA.67_156 + 16;
  vect_pa.69_257 = vect_pa.73_197 + 16;
  vect_var_.68_170 = *vect_pA.63_256;
  *vect_pa.69_257 = vect_var_.68_170;

We propagate alias info from the scalar to vector ref in
vect_create_data_ref_ptr() (in tree-vect-transform.c):

  /** (2) Add aliasing information to the new vector-pointer:
  (The points-to info (DR_PTR_INFO) may be defined later.)  **/

  tag = DR_SYMBOL_TAG (dr);
  gcc_assert (tag);

  /* If tag is a variable (and NOT_A_TAG) than a new symbol memory
 tag must be created with tag added to its may alias list.  */
  if (!MTAG_P (tag))
new_type_alias (vect_ptr, tag, DR_REF (dr));
  else
set_symbol_mem_tag (vect_ptr, tag);

Those lines do not exist on the branch. Do you take care of this somewhere
else?

Ira


-- 

irar at il dot ibm dot com changed:

   What|Removed |Added

 Status|UNCONFIRMED |NEW
 Ever Confirmed|0   |1
   Last reconfirmed|-00-00 00:00:00 |2009-01-05 13:19:53
   date||


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38721

[Bug tree-optimization/38721] [alias-improvements] vectorizer miscompiles gfortran.fortran-torture/execute/elemental.f90 at -O3

2009-01-06 Thread rguenth at gcc dot gnu dot org 


--- Comment #2 from rguenth at gcc dot gnu dot org  2009-01-06 19:42 ---
This information does no longer exist on the branch (and thus it's not
necessary
to update it).  But we are probably hitting PR38301 here, so I'll just test
fixing that bug on the branch.  Thanks for analyzing the failure!


-- 

rguenth at gcc dot gnu dot org changed:

   What|Removed |Added

  BugsThisDependsOn||38301


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38721

[Bug tree-optimization/38721] [alias-improvements] vectorizer miscompiles gfortran.fortran-torture/execute/elemental.f90 at -O3

2009-01-06 Thread rguenth at gcc dot gnu dot org 


--- Comment #3 from rguenth at gcc dot gnu dot org  2009-01-06 23:29 ---
No, the issue is that A.3 is not in gimple_addressable_vars () and thus is
considered non-aliased.  The easiest thing to do is to re-compute addressable
vars after the vectorizer.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38721

[Bug tree-optimization/38721] [alias-improvements] vectorizer miscompiles gfortran.fortran-torture/execute/elemental.f90 at -O3

2009-01-07 Thread rguenth at gcc dot gnu dot org 


--- Comment #4 from rguenth at gcc dot gnu dot org  2009-01-07 11:09 ---
Mine anyway.


-- 

rguenth at gcc dot gnu dot org changed:

   What|Removed |Added

 AssignedTo|unassigned at gcc dot gnu   |rguenth at gcc dot gnu dot
   |dot org |org
 Status|NEW |ASSIGNED
   Last reconfirmed|2009-01-05 13:19:53 |2009-01-07 11:09:14
   date||


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38721

[Bug tree-optimization/38721] [alias-improvements] vectorizer miscompiles gfortran.fortran-torture/execute/elemental.f90 at -O3

2009-01-08 Thread rguenth at gcc dot gnu dot org 


--- Comment #5 from rguenth at gcc dot gnu dot org  2009-01-08 12:29 ---
Subject: Bug 38721

Author: rguenth
Date: Thu Jan  8 12:29:46 2009
New Revision: 143185

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=143185
Log:
2009-01-07  Richard Guenther  

PR tree-optimization/38721
* tree-into-ssa.c (pass_build_ssa): Add TODO_update_address_taken.
* tree-pass.h (TODO_update_address_taken): New flag.
* tree-ssa-loop.c (tree_ssa_loop_ivopts): Update address-taken.
* tree-ssa.c (execute_update_addresses_taken): Optimize only when
optimizing.
(pass_update_address_taken): Just use TODO_update_address_taken.
* tree-flow.h (execute_update_addresses_taken): Update prototype.
* tree-cfg.c (verify_expr): Verify that stmt addresses-taken and
function addressable-vars are conservatively correct.
(verify_stmt): Initialize gsi of walk data.
* tree-inline.c (optimize_inline_calls): Execute
TODO_update_address_taken.
(tree_function_versioning): Call execute_update_addresses_taken.
* passes.c (execute_function_todo): Handle TODO_update_address_taken.
(init_optimization_passes): Remove redundant update-address-taken pass
after final inlining.
* tree-parloops.c (parallelize_loops): Call
execute_update_addresses_taken.
* tree-vectorizer.c (vectorize_loops): Likewise.

Modified:
branches/alias-improvements/gcc/ChangeLog.alias
branches/alias-improvements/gcc/passes.c
branches/alias-improvements/gcc/tree-cfg.c
branches/alias-improvements/gcc/tree-flow.h
branches/alias-improvements/gcc/tree-inline.c
branches/alias-improvements/gcc/tree-into-ssa.c
branches/alias-improvements/gcc/tree-parloops.c
branches/alias-improvements/gcc/tree-pass.h
branches/alias-improvements/gcc/tree-ssa-loop.c
branches/alias-improvements/gcc/tree-ssa.c
branches/alias-improvements/gcc/tree-vectorizer.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38721

[Bug tree-optimization/38721] [alias-improvements] vectorizer miscompiles gfortran.fortran-torture/execute/elemental.f90 at -O3

2009-01-08 Thread rguenth at gcc dot gnu dot org 


--- Comment #6 from rguenth at gcc dot gnu dot org  2009-01-08 14:59 ---
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

   What|Removed |Added

 Status|ASSIGNED|RESOLVED
 Resolution||FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38721

[Bug tree-optimization/38721] [alias-improvements] vectorizer miscompiles gfortran.fortran-torture/execute/elemental.f90 at -O3

2009-01-09 Thread rguenth at gcc dot gnu dot org 


--- Comment #7 from rguenth at gcc dot gnu dot org  2009-01-09 12:25 ---
Subject: Bug 38721

Author: rguenth
Date: Fri Jan  9 12:25:00 2009
New Revision: 143200

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=143200
Log:
2009-01-09  Richard Guenther  

Revert
2009-01-07  Richard Guenther  

PR tree-optimization/38721
* tree-into-ssa.c (pass_build_ssa): Add TODO_update_address_taken.
* tree-ssa-loop.c (tree_ssa_loop_ivopts): Update address-taken.
* tree-cfg.c (verify_expr): Verify that stmt addresses-taken and
function addressable-vars are conservatively correct.
(verify_stmt): Initialize gsi of walk data.
* tree-inline.c (optimize_inline_calls): Execute
TODO_update_address_taken.
(tree_function_versioning): Call execute_update_addresses_taken.
* passes.c (init_optimization_passes): Remove redundant
update-address-taken pass after final inlining.
* tree-parloops.c (parallelize_loops): Call
execute_update_addresses_taken.
* tree-vectorizer.c (vectorize_loops): Likewise.

Modified:
branches/alias-improvements/gcc/ChangeLog.alias
branches/alias-improvements/gcc/passes.c
branches/alias-improvements/gcc/tree-cfg.c
branches/alias-improvements/gcc/tree-inline.c
branches/alias-improvements/gcc/tree-into-ssa.c
branches/alias-improvements/gcc/tree-parloops.c
branches/alias-improvements/gcc/tree-ssa-loop.c
branches/alias-improvements/gcc/tree-vectorizer.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38721