[Bug tree-optimization/41118] Wrong dependence analysis in graphite for unrestricted pointers

[lifeng at gcc dot gnu dot org]

--- Comment #3 from lifeng at gcc dot gnu dot org  2009-09-20 10:47 ---
We should check if 2 data reference are with the same base object before
checking dependency.
This bug has been fix. Now has committed to Graphite branch.

http://gcc.gnu.org/viewcvs?view=revision&revision=151801

   * graphite-dependences.c (poly_drs_may_alias_p): Adjust definition.
   (pddr_original_scattering): Make sure 2 pdr2 in the same base object
set.
   (graphite_carried_dependence_level_k): Ditto.
   * graphite-poly.c (new_poly_dr): Add init of PDR_BASE_OBJECT_SET.
   * graphite-poly.h (struct poly_dr): Add member dr_base_object_set.
   (new_poly_dr): Adjust declaration.
   * graphite-sese-to-poly.c (free_data_refs_aux): New.
   (free_gimple_bb): Added free_data_refs_aux.
   (build_poly_dr): Add dr_base_object_set.
   (partition_drs_to_sets): New.
   (dr_same_base_object_p): New.
   (build_alias_set_for_drs): New.
   (build_base_object_set_for_drs): New.
   (build_scop_drs): Add build_base_obj_set_for_drs.
   * graphite-sese-to-poly.h: Added #define for alias set number index
and base object set index.
   * libgomp/testsuite/libgomp.graphite/force-parallel-6.c: Refine
tests.
   * libgomp/testsuite/libgomp.graphite/pr4118.c: New.


-- 

lifeng at gcc dot gnu dot org changed:

   What|Removed |Added

 CC|    |lifeng at gcc dot gnu dot
   |    |org
 Status|ASSIGNED|RESOLVED
 Resolution||FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41118

[Bug middle-end/41924] New: graphite miscompiles aermod

[lifeng at gcc dot gnu dot org]

http://www.polyhedron.com/web_images/documents/pb05.zip
the testers report this:
aermod FAILED1 fails and5 passes
[...]
Finished Testing  16 benchmarks -  15 passed, and   1 failed

--
I have reduced the aermod problem in polyhedron benchmark.
It seems that the cloog generate the wrong code for a simple
case...

---Reduced fortran code
MODULE MAIN1
 REAL , ALLOCATABLE :: HRVALD(:)
END MODULE MAIN1

SUBROUTINE VOLCALC()
 USE MAIN1
 INTEGER :: ITYP
 LOGICAL :: WETSCIM

 DO ITYP = 1 , 100
IF ( WETSCIM ) HRVALD(ITYP) = 0.0
 ENDDO
END SUBROUTINE VOLCALC
-
Equivalent C code
int hrval[101];

void volcalc()
{
 int i;
 int wetscim;

 for (i = 0; i < 100; i++)
   if (wetscim)
 hrval[i] = 0;
}
--
(gdb) set debug_generated_program (scop)

if (wetscim_3 <= -1) {
 for (scat_1=0;scat_1<=99;scat_1++) {
   S4(scat_1) ;
 }
}
if (wetscim_3 >= 1) {
 for (;;scat_1++) {
   for (git_0=0;git_0<=99;git_0++) {
 S4(git_0) ;
   }
 }
}

(gdb) set debug_scop (scop)
scop (
parameters (
p_0 -> wetscim_3(D)
)
context (
#  eq p0cst
0 3
)
pbb_4 (
conditions bb_4 (if (wetscim_3(D) != 0)
)
cases bb_4 (if (wetscim_3(D) != 0)
)
pdr_0 (write
(Data Ref:
 stmt: hrval[i_11] = 0;
 ref: hrval[i_11];
 base_object: hrval[0];
 Access function 0: {0, +, 1}_1
)
data accesses (
#  eq i0 p0  alias   sub0cst
4 6
   0  1  0  0 -1  0
   0  0  0  1  0 -1
   1 -1  0  0  0100
   1  1  0  0  0  0
)
)
domains bb_4 (
#  eq i0 p0cst
3 4
   1 -1  0 99
   1  1  0  0
   1  0 -1 -1
3 4
   1 -1  0 99
   1  1  0  0
   1  0  1 -1
)
scattering bb_4 (
#  eq s0 s1 s2 i0 p0cst
3 7
   0  1  0  0  0  0  0
   0  0  1  0 -1  0  0
   0  0  0  1  0  0  0
)
)
original_lst (
(root
 0 (loop
   0 stmt_4)))
transformed_lst (
(root
 0 (loop
   0 stmt_4)))
)


-- 
   Summary: graphite miscompiles aermod
   Product: gcc
   Version: 4.5.0
Status: UNCONFIRMED
  Severity: normal
  Priority: P3
 Component: middle-end
AssignedTo: unassigned at gcc dot gnu dot org
    ReportedBy: lifeng at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41924

[Bug middle-end/41924] graphite miscompiles aermod

[lifeng at gcc dot gnu dot org]

--- Comment #2 from lifeng at gcc dot gnu dot org  2009-11-05 02:38 ---
(In reply to comment #1)
> What are the options that you used to produce this?
> I am not able to reproduce the failing clast with -O2, nor with -O3 with the
> current graphite branch.

I produce this with -O2. It seems I use the wrong version of cloog-ppl.
I'll check this later.


-- 

lifeng at gcc dot gnu dot org changed:

   What|Removed |Added

 CC|            |lifeng at gcc dot gnu dot
   |        |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41924