[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

bin cheng  changed:

   What|Removed |Added

 Status|ASSIGNED|RESOLVED
 Resolution|--- |FIXED

--- Comment #8 from bin cheng  ---
Fixed

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

--- Comment #7 from amker at gcc dot gnu.org ---
Author: amker
Date: Thu Feb  1 12:00:57 2018
New Revision: 257291

URL: https://gcc.gnu.org/viewcvs?rev=257291=gcc=rev
Log:
Backport from mainline
2017-11-15  Bin Cheng  

PR tree-optimization/82726
PR tree-optimization/70754
* tree-predcom.c (order_drefs_by_pos): New function.
(combine_chains): Move code setting has_max_use_after to...
(try_combine_chains): ...here.  New parameter.  Sort combined chains
according to position information.
(tree_predictive_commoning_loop): Update call to above function.
(update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.

testsuite

Backport from mainline
2017-11-15  Bin Cheng  

PR tree-optimization/82726
* gcc.dg/tree-ssa/pr82726.c: New test.

Backport from mainline
2017-01-23  Bin Cheng  

PR tree-optimization/70754
* gfortran.dg/pr70754.f90: New test.

Added:
branches/gcc-6-branch/gcc/testsuite/gcc.dg/tree-ssa/pr82726.c
branches/gcc-6-branch/gcc/testsuite/gfortran.dg/pr70754.f90
Modified:
branches/gcc-6-branch/gcc/ChangeLog
branches/gcc-6-branch/gcc/testsuite/ChangeLog
branches/gcc-6-branch/gcc/tree-predcom.c

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

--- Comment #6 from amker at gcc dot gnu.org ---
Author: amker
Date: Tue Dec 19 15:25:56 2017
New Revision: 255828

URL: https://gcc.gnu.org/viewcvs?rev=255828=gcc=rev
Log:

Backport from mainline
2017-11-15  Bin Cheng  

PR tree-optimization/82726
PR tree-optimization/70754
* tree-predcom.c (order_drefs_by_pos): New function.
(combine_chains): Move code setting has_max_use_after to...
(try_combine_chains): ...here.  New parameter.  Sort combined chains
according to position information.
(tree_predictive_commoning_loop): Update call to above function.
(update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.

2017-11-15  Bin Cheng  

PR tree-optimization/82726
Revert
2017-01-23  Bin Cheng  

PR tree-optimization/70754
* tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
(reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
combined stmt before it if not NULL.
(combine_chains): Process refs reversely and compute dominance point
for root ref.

Revert
2017-02-23  Bin Cheng  

PR tree-optimization/79663
* tree-predcom.c (combine_chains): Process refs in reverse order
only for ZERO length chains, and add explaining comment.

gcc/testsuite

Backport from mainline
2017-11-15  Bin Cheng  

PR tree-optimization/82726
* gcc.dg/tree-ssa/pr82726.c: New test.


Added:
branches/gcc-7-branch/gcc/testsuite/gcc.dg/tree-ssa/pr82726.c
Modified:
branches/gcc-7-branch/gcc/ChangeLog
branches/gcc-7-branch/gcc/testsuite/ChangeLog
branches/gcc-7-branch/gcc/tree-predcom.c

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

--- Comment #5 from amker at gcc dot gnu.org ---
Author: amker
Date: Wed Nov 15 16:20:21 2017
New Revision: 254778

URL: https://gcc.gnu.org/viewcvs?rev=254778=gcc=rev
Log:
PR tree-optimization/82726
PR tree-optimization/70754
* tree-predcom.c (order_drefs_by_pos): New function.
(combine_chains): Move code setting has_max_use_after to...
(try_combine_chains): ...here.  New parameter.  Sort combined chains
according to position information.
(tree_predictive_commoning_loop): Update call to above function.
(update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.

gcc/testsuite
* gcc.dg/tree-ssa/pr82726.c: New test.

Added:
trunk/gcc/testsuite/gcc.dg/tree-ssa/pr82726.c
Modified:
trunk/gcc/ChangeLog
trunk/gcc/tree-predcom.c

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

--- Comment #4 from amker at gcc dot gnu.org ---
Author: amker
Date: Wed Nov 15 16:16:49 2017
New Revision: 254777

URL: https://gcc.gnu.org/viewcvs?rev=254777=gcc=rev
Log:
PR tree-optimization/82726
Revert
2017-01-23  Bin Cheng  

PR tree-optimization/70754
* tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
(reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
combined stmt before it if not NULL.
(combine_chains): Process refs reversely and compute dominance point
for root ref.

Revert
2017-02-23  Bin Cheng  

PR tree-optimization/79663
* tree-predcom.c (combine_chains): Process refs in reverse order
only for ZERO length chains, and add explaining comment.

Modified:
trunk/gcc/ChangeLog
trunk/gcc/tree-predcom.c

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

--- Comment #3 from amker at gcc dot gnu.org ---
Predcom uses below comparison function to make sure references in chain are
sorted in lexicographic order.  That is, inter-iteration deps are sorted in
iteration order; intra-iteration deps are sorted in dominant order.

/* Compares two drefs A and B by their offset and position.  Callback for
   qsort.  */

static int
order_drefs (const void *a, const void *b)
{
  const dref *const da = (const dref *) a;
  const dref *const db = (const dref *) b;
  int offcmp = wi::cmps ((*da)->offset, (*db)->offset);

  if (offcmp != 0)
return offcmp;

  return (*da)->pos - (*db)->pos;
}

The issue is triggered by re-association of combination, in which
re-association position is different to the original reference point.  It's
wrong to generate/replace re-associated combined reference using order of the
original combined chain.  Working on a fix.

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

Richard Biener  changed:

   What|Removed |Added

 Status|UNCONFIRMED |ASSIGNED
   Last reconfirmed||2017-10-26
Version|unknown |8.0
   Target Milestone|8.0 |---
 Ever confirmed|0   |1

--- Comment #2 from Richard Biener  ---
Confirmed.

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[amker at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

--- Comment #1 from amker at gcc dot gnu.org ---
I will investigate the issue.  It look like related to the pcom re-association
stuff.  Thanks.

[Bug tree-optimization/82726] ICE in verify_ssa during GIMPLE pass: pcom

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82726

Martin Liška  changed:

   What|Removed |Added

   Target Milestone|--- |8.0